Comparative Performance of PD-L1 Immunohistochemistry Assays: A Critical Analysis for Researchers and Drug Developers

Abstract

This article provides a comprehensive analysis of the comparative performance of various PD-L1 immunohistochemistry (IHC) assays, a critical predictive biomarker for immune checkpoint inhibitor response. We explore the foundational biology of the PD-1/PD-L1 axis and its clinical significance in immuno-oncology. The review details methodological aspects of major FDA-approved and laboratory-developed tests, including clones 22C3, 28-8, SP263, and SP142, alongside their respective scoring systems such as Tumor Proportion Score (TPS) and Combined Positive Score (CPS). We address key challenges including pre-analytical variables, inter-assay variability, and interpretation discrepancies, offering evidence-based optimization strategies. Finally, we synthesize validation requirements and comparative performance data from recent studies and meta-analyses, providing a robust framework for researchers and drug development professionals to navigate the complex PD-L1 testing landscape and advance biomarker-driven immunotherapy.

The PD-1/PD-L1 Axis: Biological Foundations and Clinical Imperative for Immunotherapy

The Role of the PD-1/PD-L1 Pathway in Immune Evasion and Cancer Surveillance

The programmed cell death protein 1 (PD-1) and its ligand (PD-L1) pathway represents a critical immune checkpoint that tumors exploit to evade host immune surveillance. Under physiological conditions, this pathway maintains self-tolerance and prevents excessive immune activation during inflammatory responses [1] [2]. However, cancer cells subvert this regulatory mechanism by overexpressing PD-L1, which engages PD-1 receptors on activated T cells, transmitting inhibitory signals that suppress T cell proliferation, cytokine production, and cytotoxic function [2] [3]. This interaction effectively creates an immunosuppressive tumor microenvironment (TME), allowing tumors to escape immune destruction—a process fundamental to cancer progression and metastasis [4] [2].

The significance of the PD-1/PD-L1 axis in oncology is underscored by the clinical success of immune checkpoint inhibitors (ICIs) that block this interaction. These therapies reinstate anti-tumor immunity by preventing PD-L1-mediated T cell suppression, leading to durable responses across multiple cancer types [1] [5]. Consequently, accurate assessment of PD-L1 expression has emerged as a crucial predictive biomarker for patient selection, driving the development and validation of various immunohistochemistry (IHC) assays for PD-L1 detection [2] [6]. This guide provides a comprehensive comparison of these analytical tools, examining their performance characteristics within the framework of comparative IHC assay research.

Mechanisms of PD-1/PD-L1-Mediated Immune Evasion

Molecular Signaling and Immune Suppression

The PD-1/PD-L1 axis suppresses T cell function through precise biochemical mechanisms. Upon PD-L1 binding to PD-1, the immunoreceptor tyrosine-based switch motif (ITSM) within the PD-1 cytoplasmic domain recruits phosphatases, primarily SHP-2 (and occasionally SHP-1) [5]. These phosphatases dephosphorylate key signaling molecules downstream of the T cell receptor (TCR), including components of the PI3K/AKT/mTOR and RAS/MEK/ERK pathways [1] [5]. This signaling inhibition results in reduced T cell proliferation, diminished cytokine production (e.g., IL-2, IFN-γ, TNF-α), impaired cytolytic activity, and ultimately promotes T cell exhaustion or apoptosis [1] [2] [3].

Cancer cells achieve persistent immune suppression through several mechanisms:

• Constitutive PD-L1 Upregulation: Driven by oncogenic signaling pathways (e.g., PI3K/AKT, JAK/STAT) and inflammatory cytokines (particularly IFN-γ) within the TME [5].
• Microenvironmental Manipulation: Secretion of immunosuppressive factors like TGF-β, IL-10, and VEGF, which further inhibit T cell function and recruit regulatory immune cells [1] [4].
• Metabolic Reprogramming: Production of lactic acid and other metabolites that create an acidic TME, directly inhibiting T cell function and promoting immunosuppressive cell populations [4].

The following diagram illustrates the core signaling mechanism through which PD-1/PD-L1 engagement inhibits T cell activation:

The Tumor Microenvironment and Immune Evasion

The immunosuppressive function of PD-L1 extends beyond tumor cells themselves. Myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs), and certain dendritic cell populations within the TME also express PD-L1, contributing to the overall immune-suppressive landscape [4]. Furthermore, metabolic competition within the TME—particularly through aerobic glycolysis leading to lactic acid accumulation—creates an acidic environment that directly inhibits T cell function and enhances PD-L1-mediated suppression [4]. This multifaceted immunosuppressive network highlights why PD-L1 has become such a critical therapeutic target and biomarker in immuno-oncology.

Comparative Performance of PD-L1 Immunohistochemistry Assays

Analytical Concordance Across Different Assays

Multiple PD-L1 IHC assays have been developed as companion diagnostics for immune checkpoint inhibitors. The substantial analytical variability between these assays presents significant challenges for clinical implementation and comparative research. A comprehensive meta-analysis of 22 studies encompassing 376 assay comparisons revealed critical differences in diagnostic accuracy when assays are used interchangeably for purposes other than their originally intended clinical application [6].

Table 1: Diagnostic Accuracy of PD-L1 IHC Assays at TPS ≥1% Cut-off (All Tissue Models)

Comparator Assay	Reference Assay	Sensitivity (95% CI)	Specificity (95% CI)	Intended Clinical Purpose
28-8 PharmDx	22C3 PharmDx	0.85 (0.82-0.88)	0.93 (0.91-0.95)	Nivolumab therapy
SP142	22C3 PharmDx	0.76 (0.71-0.81)	0.95 (0.93-0.97)	Atezolizumab therapy
SP263	22C3 PharmDx	0.91 (0.88-0.93)	0.93 (0.91-0.95)	Durvalumab therapy
73-10	22C3 PharmDx	0.97 (0.94-0.99)	0.85 (0.81-0.89)	Research use

Data adapted from meta-analysis by Røge et al. (2020) [6]

The 22C3 assay demonstrates particularly strong clinical utility in predicting response to combination chemoimmunotherapy. A prospective study of 70 NSCLC patients revealed that PD-L1 classification by the 22C3 assay showed superior correlation with therapeutic response compared to 28-8 and SP142 assays. Patients with TPS ≥50% by 22C3 had significantly longer progression-free survival, while the other assays showed no significant differences in objective response rate or survival [7].

Scoring Systems and Interpretation Criteria

PD-L1 expression is evaluated using different scoring algorithms depending on the cancer type and therapeutic context:

• Tumor Proportion Score (TPS): Percentage of viable tumor cells showing partial or complete membranous PD-L1 staining: TPS = (Number of PD-L1 positive tumor cells ÷ Total number of viable tumor cells) × 100 [2]
• Combined Positive Score (CPS): Number of PD-L1 staining cells (tumor cells, macrophages, lymphocytes) relative to the total number of viable tumor cells: CPS = (Number of PD-L1 positive tumor cells and immune cells ÷ Total number of viable tumor cells) × 100 [2]

Table 2: Clinical Cut-off thresholds for PD-L1 IHC Assays in NSCLC

Assay	Antibody Clone	ICI Drug	TPS ≥1%	TPS ≥50%	Scoring System
22C3 PharmDx	22C3	Pembrolizumab	First-line monotherapy	First-line monotherapy	TPS
28-8 PharmDx	28-8	Nivolumab	Second-line therapy	Not applicable	TPS
SP142	SP142	Atezolizumab	Not applicable	First-line combination	TC/IC
SP263	SP263	Durvalumab	Not applicable	First-line combination	TPS

Data synthesized from multiple clinical guidelines [8] [2] [6]

Interobserver and Intraobserver Variability in PD-L1 Assessment

The subjective nature of PD-L1 scoring introduces significant variability in clinical practice. A 2025 study evaluating 51 NSCLC cases scored by six pathologists demonstrated moderate interobserver agreement (Fleiss' kappa = 0.558) for TPS <1% and almost perfect agreement (Fleiss' kappa = 0.873) for TPS ≥50% [9]. Intraobserver consistency was notably higher, with Cohen's kappa ranging from 0.726 to 1.0, indicating that individual pathologists maintain consistent scoring standards over time [9]. This variability is particularly problematic for cases near critical clinical decision thresholds (TPS 1% and 50%), emphasizing the need for standardized training and quality assurance programs.

Artificial Intelligence and Digital Pathology in PD-L1 Assessment

Performance Comparison: Pathologists versus AI Algorithms

Digital pathology and artificial intelligence (AI) algorithms represent promising approaches to standardize PD-L1 scoring. However, current AI systems show inconsistent performance compared to expert pathologists. In comparative studies, AI algorithms demonstrated fair agreement (Fleiss' kappa = 0.354) for uPath software and substantial agreement (Fleiss' kappa = 0.672) for Visiopharm application at the 50% TPS cutoff when measured against median pathologist scores [9]. Notably, AI systems tend to overestimate PD-L1 positivity, particularly at lower expression thresholds, which could significantly impact patient selection for immunotherapy [9].

The following diagram outlines a typical experimental workflow for comparing pathologist and AI-based PD-L1 scoring:

Methodological Framework for Validation Studies

Robust validation of PD-L1 IHC assays requires standardized methodologies. The following experimental protocol outlines key steps for comparative performance studies:

Sample Preparation and staining:

• Collect formalin-fixed, paraffin-embedded (FFPE) tissue sections (4μm thickness) with minimum 100 viable tumor cells [9]
• Perform IHC staining using validated antibody clones (22C3, 28-8, SP263, or SP142) on automated platforms with appropriate positive and negative controls [7] [6]
• Ensure staining interpretation accounts for membranous pattern (partial or complete) with any intensity considered positive [9]

Digital Pathology and AI Analysis:

• Scan stained slides using high-resolution slide scanners (0.25 μm/pixel recommended) [9]
• Apply AI algorithms to whole slide images with manual tumor region selection when required [9]
• Implement appropriate washout periods (≥1 month) between manual and digital scoring to prevent bias [9]

Statistical Analysis:

• Calculate interobserver and intraobserver agreement using Fleiss' kappa and Cohen's kappa statistics [9]
• Determine diagnostic accuracy (sensitivity/specificity) using 2×2 contingency tables with predefined clinical thresholds [6]
• Assess correlation coefficients between different scoring methods and clinical outcomes [9] [7]

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Key Research Reagents for PD-L1 IHC Assay Development and Validation

Reagent/Material	Specifications	Research Function	Example Products
Primary Antibodies	Clones: 22C3, 28-8, SP263, SP142	PD-L1 epitope detection	Dako 22C3, Ventana SP263
IHC Detection System	Automated platforms with optimized protocols	Signal amplification and detection	Ventana Benchmark, Dako Autostainer
Tissue Controls	Cell lines with known PD-L1 expression	Assay validation and quality control	Commercial control slides
Digital Pathology System	High-resolution slide scanners	Whole slide imaging for analysis	Ventana DP200, 3DHISTECH PANORAMIC1000
AI Analysis Software	Deep learning algorithms	Automated TPS calculation	Visiopharm, uPath (Roche)
Statistical Packages	Diagnostic accuracy analysis	Data analysis and agreement metrics	R, SPSS, Stata
Etilefrine Hydrochloride	Etilefrine Hydrochloride, CAS:534-87-2, MF:C10H16ClNO2, MW:217.69 g/mol	Chemical Reagent	Bench Chemicals
Caesalmin E	Caesalmin E, MF:C26H36O9, MW:492.6 g/mol	Chemical Reagent	Bench Chemicals

Future Directions and Clinical Implications

The evolving landscape of PD-L1 detection emphasizes several critical areas for development. First, standardization of pre-analytical factors, tissue processing, and scoring criteria remains essential for reducing inter-laboratory variability [6]. Second, integrating multi-omics approaches with PD-L1 expression data may improve predictive accuracy for immunotherapy response [2] [10]. Finally, refinement of AI algorithms through larger training datasets and validation studies is necessary to achieve performance parity with expert pathologists, particularly for borderline cases [9].

Emerging technologies like liquid biopsy for soluble PD-L1 detection and multiplexed immunofluorescence for spatial analysis of the tumor immune microenvironment represent promising complementary approaches to traditional IHC [10]. However, IHC-based PD-L1 assessment remains the cornerstone for patient selection in immune checkpoint inhibitor therapy, underscoring the continued importance of rigorous comparative performance studies across different assay platforms.

For researchers conducting comparative studies, the evidence suggests that developing properly validated laboratory-developed tests for specific clinical purposes is preferable to substituting FDA-approved companion diagnostics with assays developed for different purposes [6]. This purpose-driven approach ensures that PD-L1 detection assays maintain sufficient diagnostic accuracy (sensitivity and specificity ≥90%) for their intended clinical application, ultimately optimizing patient selection and therapeutic outcomes in immuno-oncology.

Programmed Death-Ligand 1 (PD-L1) expression has emerged as the most widely adopted predictive biomarker for patient selection in immune checkpoint inhibitor (ICI) therapy. Despite its widespread clinical implementation, PD-L1 testing presents significant challenges, including assay variability, tissue heterogeneity, and imperfect predictive accuracy [11] [12]. The correlation between PD-L1 expression and treatment response varies considerably across cancer types and testing platforms, necessitating a comprehensive understanding of comparative assay performance to optimize clinical decision-making. This guide provides an objective comparison of predominant PD-L1 immunohistochemistry (IHC) assays, detailing their technical specifications, performance characteristics, and clinical utility to inform researchers and drug development professionals.

PD-L1 Signaling Pathway and Biomarker Mechanism

The biological rationale for PD-L1 as a biomarker stems from its fundamental role in immune checkpoint regulation. The PD-1/PD-L1 axis constitutes a critical immunosuppressive pathway that tumors exploit to evade host immune surveillance [13]. PD-L1 expressed on tumor cells or antigen-presenting cells binds to PD-1 receptors on T-cells, transmitting an inhibitory signal that suppresses T-cell activation, proliferation, and cytokine production [13]. This interaction effectively dampens antitumor immunity, allowing cancer cells to survive and proliferate.

Immune checkpoint inhibitors targeting the PD-1/PD-L1 axis disrupt this interaction, thereby reactivating the cytotoxic potential of T-cells and restoring antitumor immune responses [14]. The expression level of PD-L1 in tumor tissue theoretically correlates with the degree of pathway dependency, making it a mechanistically plausible biomarker for predicting ICI response [12]. However, the dynamic regulation of PD-L1 expression and the complexity of the tumor immune microenvironment contribute to the limitations observed with PD-L1 as a standalone biomarker [12] [15].

Diagram 1: PD-1/PD-L1 Signaling Pathway and Inhibitor Mechanism. This diagram illustrates how PD-L1 binding to PD-1 suppresses T-cell function and how immune checkpoint inhibitors block this interaction to restore antitumor immunity. The binding of PD-L1 on tumor cells to PD-1 on T-cells transduces inhibitory signals that suppress T-cell activation. ICIs block this interaction, preventing immune suppression and restoring T-cell-mediated cancer cell killing [13].

Comparative Analysis of PD-L1 IHC Assays

Key Commercial PD-L1 Assays

Four major PD-L1 IHC assays have been developed as companion diagnostics for specific ICIs across various cancer indications. Understanding their analytical performance and interchangeability is crucial for both clinical practice and trial design.

Table 1: Companion Diagnostic PD-L1 Assays and Their Clinical Applications

Assay Name	Primary Associated Drug	Key Cancer Indications	Scoring System	Market Share (2025)
PD-L1 IHC 22C3 pharmDx (Agilent)	Pembrolizumab	NSCLC, HNSCC, Gastric, Esophageal, Cervical	TPS, CPS	50.4% [16]
PD-L1 IHC 28-8 pharmDx (Agilent)	Nivolumab	NSCLC, Malignant Melanoma	TPS	~15% (inferred)
VENTANA PD-L1 (SP142) Assay (Roche)	Atezolizumab	TNBC, Urothelial, NSCLC	IC, TC	~15% (inferred)
VENTANA PD-L1 (SP263) Assay (Roche)	Durvalumab	Urothelial, NSCLC	TPS, CPS	~20% (inferred)

Abbreviations: TPS (Tumor Proportion Score), CPS (Combined Positive Score), IC (Immune Cell), TC (Tumor Cell), NSCLC (Non-Small Cell Lung Cancer), HNSCC (Head and Neck Squamous Cell Carcinoma), TNBC (Triple-Negative Breast Cancer)

Analytical Performance Comparison

A systematic comparability study evaluating four standardized PD-L1 assays in hepatocellular carcinoma demonstrated significant differences in analytical performance [17]. The 22C3, 28-8, and SP263 assays showed comparable sensitivity in detecting PD-L1 expression, while the SP142 assay was consistently the least sensitive across multiple scoring methods [17]. Inter-assay agreement measured by intraclass correlation coefficients was 0.646 for tumor proportion score and 0.780 for combined positive score, indicating moderate to good concordance [17].

The 22C3 assay demonstrated the strongest correlation with immune-related gene mRNA signatures, closely followed by 28-8 and SP263 assays [17]. This suggests that these three assays may provide more biologically relevant measurements of the tumor immune microenvironment compared to the SP142 assay.

Table 2: Analytical Performance Comparison of PD-L1 Assays

Performance Metric	22C3	28-8	SP263	SP142
Sensitivity in Tumor Cells	High	High	High	Low [17]
Inter-Rater Reliability (TPS)	Excellent (ICC: 0.946)	Excellent (ICC: 0.946)	Excellent (ICC: 0.946)	Good (ICC: 0.946) [17]
Inter-Rater Reliability (CPS)	Good (ICC: 0.809)	Good (ICC: 0.809)	Good (ICC: 0.809)	Lower reliability [17]
Correlation with Immune Gene Signatures	Strongest	Strong	Strong	Weaker [17]
Sample Misclassification at CPS ≥1	Low	Low	Low	Up to 18% [17]

ICC: Intraclass Correlation Coefficient

Experimental Protocols for PD-L1 Assay Comparison

Standardized Testing Methodology

The typical experimental workflow for PD-L1 assay comparison studies involves several critical steps to ensure valid and reproducible results:

• Sample Selection: Consecutive sections from surgically resected tumor specimens (e.g., hepatocellular carcinoma, NSCLC) are preferred to minimize tissue heterogeneity bias [17]. Sample sizes of approximately 55 patients provide sufficient statistical power for initial comparability assessments [17].

• Staining Protocol: Identical tumor samples are stained with four standardized PD-L1 assays (22C3, 28-8, SP142, SP263) using automated IHC platforms according to manufacturer specifications [17]. Consistent tissue processing and handling are maintained across all assays.

• Pathologist Assessment: Multiple pathologists (typically ≥5) independently evaluate PD-L1 expression using standardized scoring systems [17]. For tumor cells, the Tumor Proportion Score (TPS) is calculated as the percentage of viable tumor cells showing partial or complete membrane staining. The Combined Positive Score (CPS) accounts for both tumor cells and immune cells [14].

• Statistical Analysis: Inter-assay concordance is evaluated using intraclass correlation coefficients (ICC) for continuous scores [17]. Cohen's kappa or similar metrics assess categorical agreement at clinically relevant cutoffs (e.g., 1%, 50%). Misclassification rates are calculated relative to consensus scores.

Validation Against Molecular Signatures

To establish biological relevance, PD-L1 protein expression levels are correlated with mRNA signatures of immune-related genes using platforms like NanoString [17]. This validation step helps determine which assays most accurately reflect the underlying tumor immune microenvironment.

Advanced Biomarker Approaches Beyond Single-Parameter PD-L1

Integrated Biomarker Strategies

While PD-L1 remains the most widely validated biomarker, its limitations have prompted development of multi-parameter assessment approaches. A systematic review of 2,490 NSCLC patients across 13 studies demonstrated that combining PD-L1 with tumor-infiltrating lymphocytes (TILs) provided superior predictive value compared to either biomarker alone [15]. The hazard ratio for improved overall survival was 0.42 (95% CI: 0.31-0.56) for the combination approach, significantly outperforming PD-L1 alone (HR: 0.81) or TILs alone (HR: 0.77) [15].

Multiplex immunohistochemistry/immunofluorescence (mIHC/IF) has emerged as a promising technology, demonstrating the highest sensitivity (0.76) among biomarker testing modalities in a network meta-analysis of 5,322 patients [18]. This approach allows simultaneous evaluation of multiple immune cell populations and their spatial relationships within the tumor microenvironment.

Novel Biomarker Platforms

Table 3: Emerging Biomarker Technologies for Immunotherapy Response Prediction

Technology Platform	Mechanism	Performance Characteristics	Advantages	Limitations
Multiplex IHC/IF [18]	Simultaneous detection of multiple immune markers	Sensitivity: 0.76, Specificity: 0.65 [18]	Spatial context preservation, comprehensive immune profiling	Technical complexity, standardization challenges
Exosomal PD-L1 [19]	Detection of PD-L1 on circulating extracellular vesicles	Correlates with systemic immunosuppression	Minimally invasive, dynamic monitoring	Standardization issues, clinical interpretation evolving
Tumor Mutational Burden [18]	Quantification of total mutations	Specificity: 0.90 in gastrointestinal tumors [18]	Pan-cancer applicability, objective measurement	Cost, cutoff variability across cancer types
Gene Expression Profiling [18]	Transcriptomic signatures of immune activation	Predictive for NSCLC	Comprehensive immune status assessment	Computational complexity, validation requirements
Combined PD-L1 + TMB [18]	Dual assessment of expression and mutation burden	Sensitivity: 0.89 [18]	Improved patient selection	Increased cost and tissue requirements

Research Reagent Solutions

Table 4: Essential Research Reagents for PD-L1 Biomarker Investigation

Reagent Category	Specific Examples	Research Application	Technical Considerations
Primary Antibodies	22C3, 28-8, SP142, SP263 clones [17]	PD-L1 detection by IHC	Clone-specific epitope recognition affects staining intensity and patterns
Automated IHC Platforms	Dako Autostainer, VENTANA BenchMark [16]	Standardized staining procedures	Platform-specific antigen retrieval and detection systems
Digital Pathology Systems	Whole slide scanners with AI algorithms [16]	Quantitative image analysis	Enable consistent scoring and reduce inter-observer variability
Multiplex Detection Kits	Multiplex IHC/IF panels [18]	Simultaneous detection of multiple immune markers	Require spectral unmixing and specialized analysis software
RNA Analysis Platforms	NanoString PanCancer Immune Panel [17]	Immune-related gene expression profiling	Correlates protein expression with transcriptomic signatures
Positive Control Tissues	Cell line arrays with known PD-L1 expression [17]	Assay validation and quality control	Ensure staining consistency across experimental batches

PD-L1 remains an essential but imperfect predictive biomarker for immune checkpoint inhibitor response. The 22C3, 28-8, and SP263 assays demonstrate substantial analytical concordance, suggesting potential interchangeability in certain contexts, while the SP142 assay shows distinct performance characteristics [17]. Beyond single-parameter PD-L1 assessment, integrated approaches combining PD-L1 with TILs [15], multiplex immunofluorescence [18], or circulating biomarkers [19] show promise for improved patient stratification. As the PD-L1 biomarker testing market continues to expand—projected to grow from USD 777.2 million in 2025 to USD 1,700 million by 2035 [16]—standardization, validation, and implementation of novel technologies will be critical for advancing precision immuno-oncology. Future directions should prioritize multi-institutional validation studies and the development of clinically implementable frameworks that address both biological complexity and practical deployment challenges [11].

The advent of immune checkpoint inhibitors (ICIs) targeting the programmed cell death 1 (PD-1) and programmed death-ligand 1 (PD-L1) axis has fundamentally transformed oncology therapeutics, particularly for non-small cell lung cancer (NSCLC) and other advanced malignancies [20]. These therapies function by blocking the PD-1/PD-L1 pathway, thereby restoring the host's antitumor immunity and enabling T-cell-mediated destruction of cancer cells [21]. However, clinical benefits are not universal across all patients, creating an imperative for predictive biomarkers to identify individuals most likely to respond to treatment.

PD-L1 immunohistochemistry (IHC) has emerged as the foremost biomarker for patient selection in PD-1/PD-L1 immunotherapy [20] [22]. Its clinical implementation, however, is complicated by the development of multiple, distinct PD-L1 IHC assays, each with unique characteristics and regulatory statuses. These assays fall into two primary categories: companion diagnostics (CDx), which are essential for therapeutic decision-making as mandated by regulatory labeling, and complementary diagnostics, which provide informative data but are not strictly required for drug administration [23] [22]. This distinction carries significant implications for clinical practice, clinical trial design, and laboratory operations.

This guide provides an objective comparison of the performance characteristics of approved PD-L1 assays, detailing their analytical protocols, clinical validation data, and appropriate applications within the framework of precision immuno-oncology.

The Regulatory and Clinical Landscape of PD-L1 Assays

Defining Companion and Complementary Diagnostics

Within the context of PD-1/PD-L1 inhibitors, the regulatory classification of a diagnostic assay directly reflects its role in clinical decision-making:

• Companion Diagnostics (CDx): These assays are linked to a specific drug through their regulatory labeling, and testing is a requirement for drug use. The PD-L1 IHC 22C3 pharmDx assay (Dako) is a prominent example, serving as a companion diagnostic for pembrolizumab in NSCLC. Its use is mandatory for identifying patients eligible for treatment based on PD-L1 expression levels [23] [22].
• Complementary Diagnostics: These assays provide predictive biomarker information that can guide treatment choices but are not formally required by the drug's label. The PD-L1 IHC 28-8 pharmDx (for nivolumab) and the VENTANA PD-L1 (SP142) assay (for atezolizumab) have historically held this status for NSCLC, meaning treatment could be administered without testing, though evidence suggests a correlation between PD-L1 expression and outcomes [23].

Table 1: FDA-Approved PD-L1 Immunohistochemistry Assays and Their Status

Assay (Clone)	Platform	Primary Associated Drug(s)	Diagnostic Status	Example Indication(s)
PD-L1 IHC 22C3 pharmDx	Dako	Pembrolizumab	Companion Diagnostic	NSCLC (TPS ≥1%), HNSCC (CPS ≥1)
PD-L1 IHC 28-8 pharmDx	Dako	Nivolumab	Complementary Diagnostic	NSCLC, Melanoma
VENTANA PD-L1 (SP142)	Ventana	Atezolizumab	Complementary Diagnostic	NSCLC (TC ≥50% or IC ≥10%), TNBC
VENTANA PD-L1 (SP263)	Ventana	Durvalumab	Complementary Diagnostic	Urothelial Carcinoma

Evolving Approvals and Clinical Cutoffs

The clinical application of PD-L1 assays is governed by specific scoring systems and cutoffs established in pivotal clinical trials. These scoring methods are not uniform, adding a layer of complexity to their interpretation.

• Tumor Proportion Score (TPS): This measures the percentage of viable tumor cells displaying partial or complete membrane staining. It is used with the 22C3 and 28-8 assays for NSCLC, with common cutoffs being ≥1% and ≥50% [24] [22].
• Combined Positive Score (CPS): This calculates the number of PD-L1 staining cells (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells, multiplied by 100. It is applied for specific indications like head and neck squamous cell carcinoma (HNSCC) and cervical cancer [24].
• Immune Cell (IC) Score: The SP142 assay places greater emphasis on staining in tumor-infiltrating immune cells, requiring assessment of the percentage of tumor area occupied by PD-L1-positive immune cells [24].

The landscape of approvals is dynamic. For instance, the FDA has expanded the approval of drugs like GSK's Jemperli (dostarlimab) based on new clinical data, while occasionally rescinding previous approvals, such as the withdrawal of atezolizumab's accelerated approval for triple-negative breast cancer (TNBC) [21] [24].

Comparative Performance Analysis of PD-L1 Assays

Diagnostic Concordance and Analytical Performance

A critical question in pathology is whether the various FDA-approved PD-L1 assays are analytically equivalent and thus potentially interchangeable. Multiple studies have investigated the concordance between these assays, with findings indicating that performance is highly dependent on the clinical context and the specific clones being compared.

A landmark meta-analysis of diagnostic accuracy published in Modern Pathology evaluated interchangeability based on sensitivity and specificity for specific clinical purposes. The analysis, which included 376 assay comparisons from 22 studies, concluded that replacing an FDA-approved CDx with another assay developed for a different purpose is not advisable without proper validation. For a laboratory to substitute an approved assay, it is preferable to develop and validate a Laboratory Developed Test (LDT) for the same specific clinical purpose [6].

Recent evidence from a 2025 study in Scientific Reports focusing on clear cell renal cell carcinoma (ccRCC) underscores the challenges in concordance. This study evaluated four FDA-approved assays (22C3, 28-8, SP142, SP263) and found significant disparities, particularly with the SP142 assay, which showed remarkably lower PD-L1 positivity in immune cells (2.1%) compared to the others (~15%) [25]. The 28-8 assay demonstrated the highest pairwise concordance with other assays, while the SP142 assay was deemed an outlier.

Table 2: Comparative Diagnostic Performance from Key Studies

Assay (Clone)	Reported Sensitivity (Range)	Reported Specificity (Range)	Key Concordance Findings	Notable Limitations
22C3	Varies by study and cutoff	Varies by study and cutoff	High concordance with 28-8 and SP263 in NSCLC; κ: 0.52 with 28-8 (IC) in ccRCC [25] [6]	Gold standard for pembrolizumab, limiting cross-assay use
28-8	Varies by study and cutoff	Varies by study and cutoff	Shows highest agreement with other assays in ccRCC; moderate concordance with 22C3 & SP263 [25] [6]	Lower positivity in some tumor types (e.g., ccRCC)
SP142	Varies by study and cutoff	Varies by study and cutoff	Consistently lower positivity rates; poor concordance with other assays (κ: 0.16 with 28-8 in ccRCC) [25]	Unique scoring algorithm focusing on IC; high inter-assay variability
SP263	Varies by study and cutoff	Varies by study and cutoff	Good concordance with 22C3 and 28-8 in NSCLC and ccRCC [6]	Can show higher positivity than SP142
mIHC/IF	0.76 (95% CI: 0.57-0.89) [26]	Lower than MSI	Superior sensitivity in network meta-analysis; high predictive efficacy for NSCLC [26]	Complex methodology, not yet standardized for clinical use
MSI	Lower than mIHC/IF	0.90 (95% CI: 0.85-0.94) [26]	Highest specificity and Diagnostic Odds Ratio (DOR: 6.79); highly efficacious in GI tumors [26]	Limited to tumors with MMR deficiency

Predictive Power for Immunotherapy Response

Beyond analytical concordance, the ultimate value of a biomarker lies in its ability to predict patient response to therapy. A network meta-analysis (NMA) published in Frontiers in Immunology in 2023 compared the predictive value of various biomarkers for anti-PD-1/PD-L1 monotherapy across 49 studies [26].

This analysis revealed that multiplex immunohistochemistry/immunofluorescence (mIHC/IF) exhibited the highest sensitivity (0.76) for predicting response, suggesting it is highly effective at identifying patients who will benefit from treatment. In contrast, microsatellite instability (MSI) testing showed the highest specificity (0.90) and a high diagnostic odds ratio (6.79), making it excellent at ruling in response, particularly in gastrointestinal tumors [26].

The analysis also highlighted that combined biomarker approaches, such as PD-L1 IHC combined with tumor mutational burden (TMB), could significantly improve predictive sensitivity (0.89), underscoring the potential of multi-analyte strategies to outperform single-analyte tests like PD-L1 IHC alone [26].

Detailed Experimental Protocols for Key Studies

Protocol 1: Diagnostic Concordance Study (TMA-based)

This protocol is based on the methodology used in the 2025 ccRCC concordance study [25].

Objective: To evaluate the diagnostic concordance of four FDA-approved PD-L1 IHC assays (22C3, 28-8, SP142, SP263) in clear cell renal cell carcinoma.

Materials and Reagents:

• Formalin-Fixed Paraffin-Embedded (FFPE) tissue blocks from 286 ccRCC patients.
• Tissue Microarray (TMA) constructor.
• Automated IHC staining platforms: Dako Autostainer Link 48 (for 22C3, 28-8) and Ventana Benchmark XT or Ultra (for SP142, SP263).
• FDA-approved antibody clones and corresponding detection kits.
• Positive and negative control tissue sections.

Methodology:

• TMA Construction: Representative tumor regions were selected from donor FFPE blocks, and cylindrical cores were extracted and assembled into a recipient TMA block to allow simultaneous processing of all samples under identical conditions.
• IHC Staining: Consecutive TMA sections (4 μm thick) were cut and stained separately according to each assay's manufacturer-specified protocol, including automated staining, antigen retrieval, and detection steps. No protocol deviations were allowed.
• Pathological Evaluation: Stained slides were evaluated by qualified pathologists who were blinded to the results of other assays. Each assay was scored according to its specific evaluation criteria:
  • 22C3, 28-8, SP263: Tumor cell membrane staining was assessed.
  • SP142: Both tumor cell (TC) and tumor-infiltrating immune cell (IC) staining were evaluated, with the IC score being the percentage of tumor area occupied by stained immune cells.
• Statistical Analysis: Pairwise concordance between assays was calculated using Cohen's kappa coefficient (κ) for both tumor cell and immune cell assessments. A κ value >0.6 was generally considered to represent substantial agreement.

Protocol 2: Clinical Validation of an Alternative Antibody

This protocol is derived from the 2022 study validating the E1L3N antibody against the 22C3 companion diagnostic [27].

Objective: To assess the concordance and predictive value of the E1L3N antibody compared to the FDA-approved 22C3 assay in predicting pembrolizumab response in NSCLC.

Materials and Reagents:

• FFPE baseline tissue biopsy samples from 46 patients with unresectable, EGFR/ALK/ROS1-negative NSCLC treated with first-line pembrolizumab.
• Primary Antibodies: Rabbit anti-PD-L1 E1L3N (Cell Signaling Technology) and Mouse anti-PD-L1 22C3 (Dako).
• Automated IHC Staining System: Leica BOND-MAX.
• Bond Dewax Solution, Bond Epitope Retrieval Solution 2 (pH 9.0).

Methodology:

• Patient Selection and Sample Preparation: Patients were selected based on strict clinical criteria. Consecutive sections (4 μm) were cut from each FFPE block.
• IHC Staining (E1L3N): Staining was performed on the Leica BOND-MAX system. Steps included deparaffinization, heat-induced epitope retrieval at pH 9.0 (20 min, 100°C), incubation with E1L3N antibody, and visualization with a polymer-based detection system.
• IHC Staining (22C3): Staining for the 22C3 assay was performed according to the standard clinical protocol on the Dako platform.
• Scoring and Response Assessment: Both assay sets were scored using the Tumor Proportion Score (TPS). Patients were categorized as TPS <1%, 1-49%, or ≥50%. Treatment response was assessed radiologically every 8 weeks per RECIST 1.1 criteria. Objective Response Rate (ORR) and Progression-Free Survival (PFS) were correlated with PD-L1 expression levels from both assays.

Emerging Methodologies and Advanced Techniques

The field of PD-L1 detection is rapidly evolving beyond traditional IHC to address its limitations, such as tumor heterogeneity and the dynamic nature of PD-L1 expression.

• Liquid Biopsy for Soluble PD-L1 (sPD-L1): Detection of sPD-L1 in peripheral blood is a non-invasive alternative that allows for dynamic monitoring of PD-L1 status. sPD-L1 exists in various forms, including bound to extracellular vesicles (exosomal PD-L1) or on the surface of circulating tumor cells (CTCs) [20].
• Multiplex Immunofluorescence (mIF): This technology allows for the simultaneous detection of multiple markers (e.g., PD-L1, CD8, CD68) within the tumor microenvironment (TME). The 2023 NMA identified mIHC/IF as having superior predictive sensitivity, as it provides a more comprehensive spatial analysis of immune cell interactions [26].
• Addressing Post-Translational Modifications: PD-L1 undergoes extensive glycosylation, which can mask antibody epitopes and lead to false-negative IHC results. Novel methods incorporating enzymatic deglycosylation of tissue samples prior to IHC have been shown to improve antibody affinity and provide a more accurate quantification of PD-L1 protein levels [20].
• Combined Biomarker Approaches: As single biomarkers show limited predictive power, combining PD-L1 with other biomarkers like TMB, MSI, or gene expression profiling (GEP) is a promising strategy. Meta-analysis has confirmed that combining PD-L1 IHC with TMB significantly improves sensitivity for predicting immunotherapy response [26].

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Reagents and Materials for PD-L1 Assay Research

Item / Reagent	Function / Application	Example Specification / Notes
FFPE Tissue Sections	Substrate for IHC staining; preserves tissue morphology and antigenicity.	4-5 μm thickness; use of TMAs allows high-throughput, standardized analysis [25] [27].
FDA-Approved Antibody Clones (22C3, 28-8, SP142, SP263)	Primary antibodies for specific detection of PD-L1 in IHC.	Each clone is validated for a specific automated platform (Dako or Ventana); protocol deviations are not permitted [25] [6].
Alternative Antibodies (e.g., E1L3N)	Research-use-only (RUO) antibodies for assay development and validation.	Must be rigorously benchmarked against a clinical-grade assay for concordance and predictive value [27].
Automated IHC Staining Systems	Ensure standardized, reproducible staining conditions.	Dako Autostainer Link 48 (for 22C3/28-8); Ventana Benchmark series (for SP142/SP263); Leica BOND-MAX [25] [27].
mIHC/IF Staining Kits	Enable simultaneous detection of multiple markers on a single tissue section.	Require specialized imaging systems (e.g., multispectral scanners) and advanced bioinformatics for data analysis [26].
RNA Sequencing Kits	Profiling gene expression and identifying structural variants of PD-L1.	Used to explore mechanisms behind discordant IHC results, such as PD-L1 3'-UTR disruption [25].
2-Acetamido-3-(methylcarbamoylsulfanyl)propanoic acid	2-Acetamido-3-(methylcarbamoylsulfanyl)propanoic acid, CAS:103974-29-4, MF:C7H12N2O4S, MW:220.25 g/mol	Chemical Reagent
(S)-Lercanidipine Hydrochloride	(S)-Lercanidipine Hydrochloride, CAS:184866-29-3, MF:C36H42ClN3O6, MW:648.2 g/mol	Chemical Reagent

PD-L1 Diagnostic Decision Pathway

The following diagram illustrates the logical workflow for selecting and interpreting PD-L1 assays in clinical research, integrating key decision points regarding assay type, scoring systems, and biomarker combinations.

The landscape of predictive biomarkers for immune checkpoint inhibitor (ICI) therapy has expanded significantly beyond PD-L1 immunohistochemistry (IHC). This comparison guide objectively evaluates the performance characteristics, clinical utility, and technical considerations of PD-L1 IHC against other established biomarkers: tumor mutational burden (TMB), microsatellite instability (MSI), and mismatch repair deficiency (dMMR). We synthesize experimental data from recent network meta-analyses and clinical studies, providing structured comparisons of sensitivity, specificity, and predictive value across solid tumors. The analysis reveals that while each biomarker has distinct strengths, multiplex IHC/immunofluorescence (mIHC/IF) demonstrates superior predictive performance, and combined biomarker approaches significantly enhance prediction accuracy for anti-PD-1/PD-L1 therapy response.

The development of immune checkpoint inhibitors has revolutionized cancer treatment, creating an urgent need for reliable predictive biomarkers to identify patients most likely to respond. PD-L1 expression detected via immunohistochemistry was the first FDA-approved companion diagnostic for PD-1/PD-L1 checkpoint inhibitors but has limitations regarding variable predictive value across tumor types and methodological standardization issues [26] [28]. This has driven the exploration and validation of additional biomarkers, including tumor mutational burden (TMB), microsatellite instability (MSI), and mismatch repair deficiency (dMMR).

These biomarkers reflect different aspects of tumor immunobiology. PD-L1 IHC measures protein expression of an immune checkpoint ligand in the tumor microenvironment. TMB quantifies the total number of mutations in the tumor genome, theorized to increase neoantigen production and immunogenicity. MSI and dMMR represent functional deficits in DNA repair machinery that lead to hypermutation. Understanding their comparative performance characteristics, technical requirements, and clinical applications is essential for optimizing treatment selection and advancing personalized cancer immunotherapy [26] [29].

This review provides a comprehensive comparison of these biomarkers within the broader context of comparative performance of immunohistochemistry assays for PD-L1 detection research, presenting structured experimental data and methodological protocols to guide researchers and drug development professionals.

Biomarker Definitions and Biological Basis

PD-L1 Expression

Programmed Death-Ligand 1 (PD-L1) is an immune checkpoint protein expressed on tumor cells and various immune cells. Its interaction with PD-1 receptors on T cells inhibits T-cell activation, enabling tumor immune evasion. PD-L1 IHC detects the presence of this protein in tumor tissue, with various scoring systems including Tumor Proportion Score (TPS) and Combined Positive Score (CPS) [28]. The biological rationale suggests tumors expressing PD-L1 may be more dependent on this pathway for immune escape and thus more susceptible to PD-1/PD-L1 blockade.

Tumor Mutational Burden (TMB)

TMB measures the total number of somatic mutations per megabase of interrogated genomic sequence. High TMB is hypothesized to increase neoantigen formation, enhancing tumor immunogenicity and T-cell recognition. When checkpoint inhibition is applied, these highly mutated tumors may generate more robust anti-tumor immune responses [29]. TMB is typically assessed using next-generation sequencing (NGS) panels or whole-exome sequencing.

Mismatch Repair Deficiency (dMMR) and Microsatellite Instability (MSI)

dMMR results from functional defects in DNA mismatch repair proteins (MLH1, MSH2, MSH6, PMS2), typically detected by IHC showing loss of protein expression. MSI represents the phenotypic consequence of dMMR—widespread insertion/deletion mutations at microsatellite regions throughout the genome, detected by PCR or NGS [30] [31]. This hypermutated state generates abundant frameshift-derived neoantigens, creating highly immunogenic tumors particularly responsive to immune checkpoint blockade [30].

The following diagram illustrates the biological relationships between these biomarkers and their connection to immunotherapy response:

Diagram 1: Biological relationships between predictive biomarkers and their detection methods. This map illustrates how dMMR drives MSI and high TMB, leading to neoantigen generation and T-cell infiltration, which subsequently induces PD-L1 expression as an immune evasion mechanism. The dotted lines indicate detection methodologies for each biomarker.

Comparative Performance Data

Diagnostic Accuracy Across Biomarkers

A recent network meta-analysis (NMA) comparing different predictive biomarker testing assays for PD-1/PD-L1 checkpoint inhibitors provides comprehensive performance data across 49 studies covering 5,322 patients [26] [18]. The analysis evaluated seven biomarker approaches: PD-L1 IHC, TMB, gene expression profiling (GEP), MSI, multiplex IHC/immunofluorescence (mIHC/IF), other IHC/hematoxylin-eosin staining, and combined assays.

Table 1: Diagnostic accuracy of predictive biomarkers for anti-PD-1/PD-L1 therapy response

Biomarker	Sensitivity (95% CI)	Specificity (95% CI)	Diagnostic Odds Ratio (95% CI)	Superiority Index
mIHC/IF	0.76 (0.57-0.89)	0.67 (0.47-0.82)	5.09 (1.35-13.90)	2.86
MSI	0.44 (0.30-0.60)	0.90 (0.85-0.94)	6.79 (3.48-11.91)	2.59
PD-L1 IHC	0.54 (0.45-0.62)	0.76 (0.68-0.83)	3.83 (2.56-5.56)	1.98
TMB	0.45 (0.35-0.56)	0.77 (0.68-0.84)	2.71 (1.69-4.17)	1.56
GEP	0.63 (0.46-0.77)	0.65 (0.47-0.79)	3.21 (1.26-7.14)	1.82
Combined PD-L1 IHC + TMB	0.89 (0.82-0.94)	0.53 (0.42-0.64)	7.94 (4.20-14.49)	3.52

The data reveal that mIHC/IF exhibited the highest sensitivity (0.76), while MSI showed the highest specificity (0.90) and diagnostic odds ratio (6.79). Combined PD-L1 IHC with TMB demonstrated markedly improved sensitivity (0.89) compared to either biomarker alone [26].

Tumor-Type Specific Performance

Biomarker performance varies significantly across cancer types, reflecting differences in tumor immunobiology and oncogenic drivers.

Table 2: Biomarker performance across different tumor types

Tumor Type	Optimal Biomarker(s)	Key Findings	Supporting Evidence
Non-small cell lung cancer (NSCLC)	mIHC/IF, Other IHC&HE	mIHC/IF demonstrated high predictive efficacy	[26]
Gastrointestinal tumors	PD-L1 IHC, MSI	MSI shows high specificity (0.90) and DOR (6.79)	[26]
Colorectal cancer	dMMR/MSI	High prevalence of dMMR (8.7-26.8%) and MSI (8.5-21.9%)	[32]
Endometrial cancer	dMMR/MSI, TMB	High prevalence of dMMR (8.7-26.8%) and MSI (8.5-21.9%)	[32]
Esophageal squamous cell carcinoma	PD-L1, TMB	54% PD-L1+, 57% TMB-H, but only 1% MSI-H	[33]
Anal squamous cell carcinoma	PD-L1	64.25% expressed PD-L1; PD-L1-high associated with longer treatment duration	[34]
Cervical, Bladder/Urothelial, Lung, Skin cancers	TMB	Low dMMR/MSI prevalence (<5%) but high TMB-H (23.7-52.6%)	[32]

Pan-Cancer Prevalence

A comprehensive scoping review and meta-analysis of 3890 papers provides population-level prevalence data for these biomarkers [32]:

Table 3: Pan-cancer prevalence of predictive biomarkers

Biomarker	Pooled Overall Prevalence	High Prevalence Cancers	Low Prevalence Cancers
dMMR	2.9%	Endometrial (8.7-26.8%), Colorectal (8.7-26.8%), Small Bowel (8.7-26.8%), Gastric (8.7-26.8%)	Cervical, Esophageal, Bladder/Urothelial, Lung, Skin (<5%)
MSI	2.7%	Endometrial (8.5-21.9%), Colorectal (8.5-21.9%), Small Bowel (8.5-21.9%), Gastric (8.5-21.9%)	Cervical, Esophageal, Bladder/Urothelial, Lung, Skin (<5%)
High TMB (≥10 mut/Mb)	14.0%	Cervical (23.7-52.6%), Esophageal (23.7-52.6%), Bladder/Urothelial (23.7-52.6%), Lung (23.7-52.6%), Skin (23.7-52.6%)	Other cancer types (generally <5%)

Methodological Approaches and Experimental Protocols

PD-L1 IHC Testing Protocols

Standard Protocol:

• Tissue Processing: Formalin-fixed paraffin-embedded (FFPE) tumor specimens sectioned at 4-5μm [33]
• Staining Platform: Ventana BenchMark ULTRA automated staining platform [33]
• Antibody Clones: SP263, 22C3, SP142, 28-8 (various companion diagnostics)
• Epitope Retrieval: Heat-induced epitope retrieval with Cell Conditioning Solution (CC1, Tris-EDTA buffer, pH 8.0) for 64 minutes at 95°C [33]
• Detection: OptiView DAB IHC Detection Kit with hematoxylin counterstaining [33]
• Scoring Systems: Tumor Proportion Score (TPS) or Combined Positive Score (CPS) with cutoff values varying by cancer type and therapeutic context

TMB Assessment Methods

NGS-Based Approaches:

• Platforms: Illumina NextSeq or NovaSeq platforms [34]
• Gene Panels: FDA-approved panels include FoundationOne CDx (324 genes, 0.8 Mb), MSK-IMPACT (468 genes, 1.14 Mb) [29]
• Sequencing Depth: Minimum 500× coverage with analytic sensitivity of 5% [34]
• Mutation Types Included: Non-synonymous missense, nonsense, in-frame insertion/deletion, and frameshift mutations [29]
• Calculation: Number of mutations divided by the size of the genomic territory, reported as mutations per megabase (mut/Mb)
• Thresholds: TMB-H typically defined as ≥10 mut/Mb, though optimal cutpoints vary by cancer type [29]

dMMR/MSI Testing Methodologies

Dual-Modality Approach:

• dMMR IHC Protocol:
  • FFPE sections stained for four MMR proteins (MLH1, MSH2, MSH6, PMS2) using automated platforms
  • Interpretation: Loss of nuclear expression in tumor cells with positive internal control
  • Patterns: Classical (MLH1/PMS2 or MSH2/MSH6 loss), non-classical, or unusual (focal/subclonal) patterns [31]
• MSI PCR Protocol:
  • DNA extraction from macro-dissected tumor and normal areas
  • PCR amplification using five mononucleotide markers (BAT-25, BAT-26, MONO-27, NR-21, NR-24) plus pentanucleotide controls
  • Fragment analysis by capillary electrophoresis
  • Classification: MSI-H (≥2 unstable markers), MSI-L (1 unstable marker), MSS (no unstable markers) [31]

The following diagram illustrates the typical testing workflow for these biomarkers:

Diagram 2: Biomarker testing workflow from tumor sample to result interpretation. This flowchart illustrates the parallel and sometimes overlapping methodologies for different biomarkers, highlighting how IHC, NGS, and PCR approaches generate complementary predictive information from tumor samples.

Technical Challenges and Methodological Considerations

Concordance Between dMMR and MSI Testing

Despite theoretical equivalence, significant discrepancies exist between dMMR IHC and MSI PCR results. A large comparative study of 703 cases found a 19.3% overall discrepancy rate, with particularly high rates (60.9%) in dMMR versus MSI-high comparisons [31]. This discordance appears independent of tumor types and not fully explained by technical factors like tumor percentage.

Potential contributors to discordance include:

• Unusual dMMR patterns: Focal/subclonal or heterogeneous protein loss
• Variant MMR genes: Mutations causing functional impairment without complete protein loss
• Technical factors: Antibody sensitivity, staining interpretation, preanalytical variables
• Marker performance variability: In non-colorectal cancers, NR-21 and NR-24 markers show lower sensitivity (67-73%) [31]

TMB Measurement Standardization

Multiple challenges complicate TMB measurement standardization:

• Panel size variability: Commercially available panels cover 0.80-2.40 Mb, affecting precision [29]
• Gene content differences: Varying genes and genomic regions covered
• Bioinformatic pipelines: Laboratory-specific algorithms and filters
• Mutation types included: Inclusion/exclusion of synonymous mutations
• Cutpoint determination: Cancer-type-specific versus pan-cancer thresholds

The coefficient of variation of TMB estimation decreases inversely with both the square root of panel size and TMB level—halving the CV requires a four-fold increase in panel size [29].

PD-L1 IHC Limitations

PD-L1 testing faces several methodological challenges:

• Spatial heterogeneity: Variable expression within tumors and between primary/metastatic sites
• Temporal dynamics: Expression changes over time and with prior therapies
• Scoring variability: Inter-observer reproducibility issues and different scoring systems
• Platform/assay differences: Multiple FDA-approved assays with potentially different performance characteristics
• Dynamic biology: Inducible nature of PD-L1 expression influenced by microenvironmental factors

The Researcher's Toolkit: Essential Reagents and Platforms

Table 4: Key research reagent solutions for predictive biomarker analysis

Category	Specific Products/Platforms	Application	Key Features
IHC Platforms	Ventana BenchMark ULTRA	Automated PD-L1 and MMR protein staining	Standardized staining with FDA-approved protocols
PD-L1 Antibody Clones	SP263, 22C3, SP142, 28-8	PD-L1 protein detection	Companion diagnostics for specific therapeutics
MMR IHC Antibodies	MLH1 (M1), MSH2 (G219-1129), MSH6 (SP93), PMS2 (A16-4)	dMMR detection	Ventana ready-to-use monoclonal antibodies
NGS Panels	FoundationOne CDx, MSK-IMPACT, TSO500	TMB measurement, MSI detection	Comprehensive genomic profiling with validated TMB calculation
MSI PCR Kits	Promega MSI Analysis System Version 1.2	MSI status determination	Five mononucleotide markers with pentanucleotide controls
DNA Extraction Kits	Qiagen AllPrep DNA/RNA FFPE Kit	Nucleic acid isolation from FFPE	Simultaneous DNA/RNA extraction from challenging samples
Analysis Software	quanTIseq, Immune Cell Abundance	Tumor microenvironment quantification	Computational deconvolution of immune cell populations
Propiverine Hydrochloride	Propiverine Hydrochloride, CAS:54556-98-8, MF:C23H30ClNO3, MW:403.9 g/mol	Chemical Reagent	Bench Chemicals
Rosiglitazone-d3	Rosiglitazone-d3 | Stable Isotope | For Research Use	Rosiglitazone-d3, a deuterated internal standard. Essential for accurate LC-MS/MS quantification in metabolism studies. For Research Use Only. Not for human use.	Bench Chemicals

This comprehensive comparison reveals that each predictive biomarker for ICI response has distinct strengths and limitations. PD-L1 IHC provides direct measurement of the therapeutic target but suffers from heterogeneity and dynamic regulation. TMB offers a quantitative measure of tumor immunogenicity with pan-cancer applicability but requires standardization. dMMR/MSI identifies a biologically distinct tumor subset with exceptional response rates but limited prevalence across cancers.

The emerging paradigm favors integrated biomarker approaches rather than reliance on single markers. Combined PD-L1 IHC with TMB significantly enhances sensitivity [26], while multiplex IHC/IF technologies provide spatial context that improves predictive power. Future directions should focus on standardizing measurement approaches, validating combinatorial biomarker algorithms, and developing novel methodologies that capture the complexity of tumor-immune interactions across diverse cancer types.

Clinical Consequences of Testing Variability on Patient Selection and Treatment Outcomes

Programmed Death-Ligand 1 (PD-L1) expression testing via immunohistochemistry (IHC) serves as a critical predictive biomarker for immune checkpoint inhibitor (ICI) therapy in non-small cell lung cancer (NSCLC). However, substantial variability in PD-L1 testing methodologies and interpretation significantly impacts patient selection and subsequent treatment outcomes. This variability stems from multiple factors including pre-analytical conditions, choice of IHC assays, pathologist interpretation, and tumor biological characteristics. Understanding these sources of variability and their clinical consequences is essential for optimizing personalized immunotherapy approaches. This guide systematically compares the performance of different PD-L1 assessment methods, evaluates their impact on treatment decisions, and provides evidence-based recommendations for reducing variability in clinical practice.

Biological Context: PD-L1 Signaling and Immune Evasion

The PD-1/PD-L1 pathway represents a critical immune checkpoint mechanism that tumors exploit to evade host immune surveillance. PD-L1, expressed on tumor cells and resident immune cells, binds to PD-1 receptors on activated T-cells, thereby inhibiting T-cell receptor signaling and suppressing cytotoxic T-cell function [35]. This interaction leads to reduced T-cell proliferation, decreased cytokine production, and diminished cytotoxic activity, ultimately facilitating tumor immune escape [35]. Immune checkpoint inhibitors targeting this pathway—including anti-PD-1 and anti-PD-L1 antibodies—block this interaction, restoring antitumor immunity and enabling T-cell-mediated tumor cell killing [36].

The following diagram illustrates the PD-1/PD-L1 signaling pathway and mechanism of immune checkpoint inhibition:

Diagram 1: PD-1/PD-L1 signaling pathway and immune checkpoint inhibition mechanism. Tumor cells express PD-L1 which binds to PD-1 on T-cells, leading to T-cell inhibition. Immune checkpoint inhibitors block this interaction, restoring T-cell function.

Pre-Analytical and Analytical Factors

Multiple technical factors contribute to variability in PD-L1 testing results. Pre-analytical conditions including specimen age significantly impact PD-L1 detectability, with longer storage times associated with reduced detection rates [37]. A comprehensive meta-analysis of 92 studies demonstrated that PD-L1 detectability declines with increasing specimen age, while consistency improves when data are pooled from multiple laboratories [37]. Additionally, different IHC assays utilizing various antibody clones (e.g., 22C3, SP263, 28-8, SP142) demonstrate varying sensitivities and specificities, leading to interpretation discrepancies particularly at lower expression thresholds [38] [37].

Tumor Biological Factors

Biological characteristics of tumors introduce significant variability in PD-L1 assessment. Tumor heterogeneity—both spatial and temporal—represents a major challenge, with studies demonstrating only approximately 30% concordance in PD-L1 expression between paired primary tumors and metastatic lymph nodes [39]. Intrapatient variation in PD-L1 expression can be substantial, with major increases (ΔTPS ≥ +50%) and decreases (ΔTPS ≤ -50%) observed in 9.7% and 8.0% of cases, respectively [40]. Furthermore, intervening ICI therapy is associated with decreased PD-L1 expression, while acquired copy number losses of CD274, PDCD1LG2, and JAK2 genes are strongly associated with major decreases in PD-L1 expression [40].

Interpretation and Scoring Variability

Subjective interpretation of PD-L1 expression represents a significant source of variability. Studies demonstrate moderate interobserver agreement among pathologists at the TPS <1% cutoff (Fleiss' kappa 0.558) and almost perfect agreement at TPS ≥50% (Fleiss' kappa 0.873) [9]. Intraobserver consistency is generally higher, with Cohen's kappa ranging from 0.726 to 1.0 [9]. This variability has direct clinical implications, as patients with multiple PD-L1 assessments before ICI therapy showing all samples with PD-L1 ≥1% achieved improved objective response rates and progression-free survival compared to cases with discordant results (at least one sample with PD-L1 <1% and another with PD-L1 ≥1%) [40].

Table 1: Key Factors Contributing to PD-L1 Testing Variability

Factor Category	Specific Factors	Impact on Variability	Clinical Consequences
Pre-analytical	Specimen age, fixation methods, specimen type (biopsy vs. resection)	Reduced PD-L1 detectability with longer storage; concordance issues between sample types	False-negative results leading to potential exclusion from beneficial therapy
Analytical	IHC assay platform, antibody clone (22C3, SP263, 28-8, SP142), staining platforms	Different sensitivities and specificities; inter-assay discordance particularly at low expression levels	Inaccurate patient stratification and potential treatment assignment errors
Tumor Biological	Spatial heterogeneity, temporal changes, genomic alterations (CD274, JAK2)	Major expression changes (ΔTPS ≥ ±50%) in ~18% of cases; heterogeneity-driven sampling errors	Discordant treatment responses; acquired resistance mechanisms
Interpretation	Pathologist experience, interobserver variability, scoring criteria	Moderate agreement at TPS <1% (kappa 0.558); better agreement at TPS ≥50% (kappa 0.873)	Inconsistent treatment thresholds affecting patient selection

Comparative Performance of PD-L1 Assessment Methods

Pathologist versus Artificial Intelligence Scoring

The emergence of artificial intelligence (AI) algorithms for PD-L1 scoring presents both opportunities and challenges for standardizing assessment. When comparing pathologists to AI algorithms, pathologists demonstrate higher consistency at critical Tumor Proportion Score (TPS) cutoffs in NSCLC [9]. In a comparative study of 51 NSCLC cases, pathologists showed moderate interobserver agreement for TPS <1% (Fleiss' kappa 0.558) and almost perfect agreement for TPS ≥50% (Fleiss' kappa 0.873) [9]. Intraobserver consistency was high, with Cohen's kappa ranging from 0.726 to 1.0 [9].

Comparisons between AI algorithms and median pathologist scores showed fair agreement for uPath software (Fleiss' kappa 0.354) and substantial agreement for the Visiopharm application (Fleiss' kappa 0.672) at the 50% TPS cutoff [9]. These results indicate that while AI tools show promise, they currently cannot fully replace expert human evaluation, particularly in critical clinical decision-making contexts requiring refinement to match pathologist reliability [9].

Novel Computational Approaches

Quantitative continuous scoring (QCS) approaches represent an innovative alternative to traditional semi-quantitative assessment. PD-L1 QCS utilizes computer vision systems for granular cell-level quantification of PD-L1 staining intensity in digitized whole slide images [41]. This methodology captures the percentage of tumor cells with medium to strong staining intensity (PD-L1 QCS-PMSTC) and classifies patients with ≥0.575% as biomarker-positive [41].

In the MYSTIC trial, visual PD-L1 scoring (TPS ≥50%) resulted in a hazard ratio of 0.69 (CI 0.46-1.02) with a 29.7% prevalence of biomarker-positive patients for durvalumab versus chemotherapy [41]. With PD-L1 QCS-PMSTC, a similar hazard ratio of 0.62 (CI 0.46-0.82) was obtained with an increased prevalence of 54.3% [41]. This demonstrates that quantitative approaches can identify broader patient populations who may benefit from ICI therapy while maintaining similar treatment effects.

Circulating Tumor Cell Analysis

Liquid biopsy approaches utilizing circulating tumor cells (CTCs) offer an alternative to tissue-based PD-L1 assessment that captures tumor heterogeneity across multiple metastatic sites. Quantitative microscopic evaluation of PD-L1 and HLA I expression on CTCs from NSCLC patients demonstrates heterogeneity in expression patterns and shows promising clinical value in predicting progression-free survival in response to PD-L1 targeted therapies [39].

The analytical validation of exclusion-based sample preparation technology for CTC analysis demonstrates high precision and accuracy, confirming compatibility for clinical laboratory implementation [39]. This approach addresses spatial and temporal heterogeneity limitations of tissue biopsies and enables serial monitoring of PD-L1 expression dynamics during treatment.

Table 2: Comparison of PD-L1 Assessment Method Performance

Assessment Method	Agreement/Performance Metrics	Advantages	Limitations
Pathologist Visual Scoring	Interobserver: TPS <1% (kappa 0.558), TPS ≥50% (kappa 0.873)Intraobserver: kappa 0.726-1.0 [9]	Clinical standard; expert interpretation; handles complex morphology	Subjectivity; moderate agreement at low expression levels; fatigue
AI Algorithm (uPath)	Fair agreement with pathologists (kappa 0.354) at TPS ≥50% [9]	Quantitative; rapid processing; reduces labor intensiveness	Lower agreement than pathologists; requires manual tumor area selection
AI Algorithm (Visiopharm)	Substantial agreement with pathologists (kappa 0.672) at TPS ≥50% [9]	Better agreement profile; automated analysis potential	Still requires refinement for clinical decision-making
Quantitative Continuous Scoring (QCS)	HR 0.62 vs chemotherapy; identifies 54.3% as biomarker-positive vs 29.7% with visual scoring [41]	Continuous scoring; identifies more potential responders; granular intensity measurement	Computational complexity; requires validation across platforms
CTC-Based Analysis	Heterogeneous expression patterns; predicts PFS to PD-L1 therapy [39]	Captures spatial heterogeneity; enables serial monitoring; minimally invasive	Technical challenges in rare cell capture; not yet standardized

Clinical Consequences of Testing Variability

Impact on Patient Selection and Treatment Patterns

PD-L1 testing variability directly influences patient selection for ICI therapy and subsequent treatment patterns. Real-world evidence from 507 patients with metastatic NSCLC demonstrated increasing PD-L1 testing rates from 86% in 2017 to 100% in 2020, reflecting growing recognition of its clinical importance [42]. However, treatment selection varied significantly based on PD-L1 expression levels and histomolecular subtypes.

In patients with nonsquamous NSCLC without actionable genomic alterations, ICI-chemotherapy combinations were the most common first-line regimens except in the PD-L1 ≥50% category, where ICI monotherapy was most frequently administered [42]. Use of chemotherapy decreased while ICI-chemotherapy combinations increased from 2017 to 2020 across all histomolecular groups [42]. These patterns demonstrate how PD-L1 expression levels directly guide therapeutic decisions, with testing variability potentially leading to substantial deviations from optimal treatment pathways.

Survival Outcomes Based on Testing and Treatment Selection

Testing variability ultimately impacts clinical outcomes, including overall survival (OS). For all patients with metastatic NSCLC in the real-world study, median OS was 25.0 months (95% CI, 19.1-28.3), with significant variation by histomolecular cohort: 14.3 months for squamous NSCLC, 25.3 months for nonsquamous NSCLC with no actionable genomic alteration, not reached for KRAS G12C-mutated NSCLC, and 27.7 months for nonsquamous NSCLC with other genomic alterations [42].

The clinical consequence of PD-L1 expression variation is further highlighted by studies showing that among patients with multiple PD-L1 assessments before ICI, cases where all samples had PD-L1 ≥1% achieved improved objective response rate and progression-free survival compared to cases with discordant results [40]. Additionally, when the most proximal sample before ICI therapy showed PD-L1 ≥1%, patients had longer median PFS compared to cases where the most proximal sample was PD-L1 <1% [40]. This underscores the critical impact of temporal testing variability on treatment outcomes.

Standardized Experimental Protocols

Tissue-Based PD-L1 IHC Protocol

The following protocol represents a standardized approach for tissue-based PD-L1 immunohistochemical analysis:

Sample Preparation: Use freshly cut 4-μm-thick formalin-fixed paraffin-embedded (FFPE) sections from tissue specimens containing at least 100 tumor cells [9]. Ensure appropriate fixation times (6-72 hours in 10% neutral buffered formalin) to prevent antigen degradation.

Staining Procedure: Apply validated PD-L1 antibody clones (e.g., SP263, 22C3, 28-8) according to manufacturer protocols on automated staining platforms such as the BenchMark ULTRA [9]. Include appropriate positive and negative controls with each staining run.

Scoring Methodology: Evaluate PD-L1 staining only on tumor cells, considering any intensity of either partial or complete membranous staining as positive [9]. Record the percentage of positively stained tumor cells as follows: 0%, 1%, 5%, 10%, and up to 100% in 10% increments [9]. For digital pathology, scan slides with resolution of at least 0.25 μm/pixel on slide scanners such as PANORAMIC1000 or Ventana DP200 [9].

Quality Assurance: Implement regular proficiency testing and laboratory comparison programs to minimize inter-laboratory variability. Adhere to CAP-PLQC guidelines for validation and ongoing quality control.

Quantitative Continuous Scoring Protocol

For quantitative continuous scoring of PD-L1 expression:

Image Acquisition: Scan PD-L1-stained slides at high resolution (minimum 0.25 μm/pixel) using whole slide scanners. Ensure uniform focus and illumination across entire slide [41].

Image Analysis: Apply computer vision systems for granular cell-level quantification of PD-L1 staining intensity. Define positive cells as having PD-L1 membrane staining intensity ≥40 (on a 0-255 scale) [41]. Calculate the percentage of tumor cells meeting this intensity threshold.

Biomarker Classification: Classify samples as biomarker-positive where >0.575% of tumor cells demonstrate medium to strong staining intensity (PD-L1 QCS-PMSTC) [41]. This threshold optimizes identification of patients likely to benefit from ICI therapy.

Validation: Compare QCS results with pathologist-derived tumor proportion scores at ≥1% and ≥50% cutoffs to ensure concordance. Validate against clinical outcomes from relevant trials when possible.

The following workflow diagram illustrates the standardized protocol for PD-L1 assessment:

Diagram 2: Standardized workflow for PD-L1 immunohistochemical analysis. The process includes pre-analytical, analytical, and post-analytical phases to ensure consistent and reliable PD-L1 assessment.

The Scientist's Toolkit: Essential Research Reagents and Platforms

Table 3: Key Research Reagent Solutions for PD-L1 Assessment

Reagent/Platform	Manufacturer	Function	Application Notes
VENTANA PD-L1 (SP263) Assay	Ventana Medical Systems/Roche	IHC detection of PD-L1 expression	Used on BenchMark ULTRA platform; validated for NSCLC [9]
pharmDx 22C3 Anti-PD-L1	Agilent/Dako	IHC detection of PD-L1 expression	FDA-approved companion diagnostic for pembrolizumab [38]
uPath PD-L1 Software	Roche	Digital image analysis for PD-L1 scoring	IVDD-certified; requires manual tumor area selection [9]
PD-L1 Lung Cancer TME App	Visiopharm	AI-based digital scoring of PD-L1	Research-use-only; shows substantial agreement with pathologists [9]
ExtractMax System	Gilson and Salus	Automated circulating tumor cell isolation	Enables high-yield CTC capture for liquid biopsy approaches [39]
PANORAMIC1000 Slide Scanner	3DHISTECH	Whole slide image digitization	0.25μm/pixel resolution for high-quality digital pathology [9]
Ventana DP200 Slide Scanner	Roche Diagnostics	Whole slide image digitization	Compatible with uPath software platform [9]
Ciprofloxacin-d8	Ciprofloxacin-d8 Hydrochloride Hydrate	Ciprofloxacin-d8 HCl hydrate, a deuterium-labeled internal standard for quantitative LC-MS/MS analysis of ciprofloxacin in research samples. For Research Use Only (RUO). Not for human use.	Bench Chemicals
Temozolomide-d3	Temozolomide-d3, CAS:208107-14-6, MF:C6H6N6O2, MW:197.17 g/mol	Chemical Reagent	Bench Chemicals

PD-L1 testing variability significantly impacts patient selection and treatment outcomes in NSCLC immunotherapy. Multiple factors contribute to this variability, including pre-analytical conditions, assay selection, tumor biological characteristics, and interpretation differences. The clinical consequences are substantial, affecting treatment choices and ultimately survival outcomes. Emerging approaches including artificial intelligence algorithms, quantitative continuous scoring, and liquid biopsy methods offer potential pathways to reduce variability and improve patient stratification. Standardization of testing protocols, implementation of quality assurance programs, and adoption of validated computational approaches will be essential to minimize variability and optimize personalized immunotherapy approaches. Future research should focus on integrating multiple biomarkers including tumor mutational burden and HLA expression to complement PD-L1 testing and improve predictive accuracy for immune checkpoint inhibitor response.

PD-L1 Assay Methodologies: From Automated Platforms to Scoring Algorithms

The advent of immune checkpoint inhibitors (ICIs) has revolutionized cancer treatment, harnessing the body's immune system to combat malignant cells. The interaction between programmed death 1 (PD-1) on T cells and its ligand (PD-L1) on tumor cells constitutes a critical mechanism for immune escape, making the PD-1/PD-L1 pathway a prime therapeutic target [43]. Accurate assessment of PD-L1 expression levels through immunohistochemistry (IHC) has become an essential component of precision oncology, enabling identification of patients most likely to benefit from ICI therapy [44].

Four PD-L1 IHC assays—22C3, 28-8, SP263, and SP142—have received approval from the U.S. Food and Drug Administration (FDA) as companion or complementary diagnostics for various ICIs across multiple cancer types [25] [44]. These assays were developed independently, utilizing different antibody clones, staining platforms, and scoring algorithms, which has resulted in challenges regarding their concordance and interchangeability in clinical practice [25] [45]. This comparison guide provides a detailed, evidence-based analysis of these four major assays, focusing on their technical specifications, analytical performance, and clinical utility across different tumor types to inform researchers, scientists, and drug development professionals.

Assay Specifications and Scoring Systems

The four FDA-approved PD-L1 assays employ distinct antibody clones, detection platforms, and scoring criteria, leading to differences in PD-L1 positivity rates and interpretation.

Table 1: Technical Specifications of FDA-Approved PD-L1 Assays

Assay Clone	Platform	Primary Scoring Method	Key FDA-Approved Indications	Tumor Cell vs. Immune Cell Staining Emphasis
22C3	Dako Link 48	Tumor Proportion Score (TPS)	NSCLC, HNSCC, GCC, ESCCC, UC	Primarily tumor cells
28-8	Dako Link 48	Tumor Proportion Score (TPS)	NSCLC, RCC, MCC	Primarily tumor cells
SP263	Ventana Benchmark	Tumor Cell Staining	NSCLC, UC	Balanced tumor and immune cells
SP142	Ventana Benchmark	Immune Cell (IC) Score	TNBC, UC	Emphasis on immune cells

The 22C3 pharmDx assay employs the Tumor Proportion Score (TPS), defined as the percentage of viable tumor cells showing partial or complete membranous PD-L1 staining relative to all viable tumor cells [45] [46]. This assay serves as a companion diagnostic for pembrolizumab in multiple malignancies including non-small cell lung cancer (NSCLC) [44].

The 28-8 pharmDx assay similarly utilizes TPS scoring and is approved as a complementary diagnostic for nivolumab in NSCLC and other cancers [25]. It shares the Dako platform with the 22C3 assay but employs a different antibody clone.

The SP263 assay on the Ventana platform assesses the percentage of tumor cells with any membranous PD-L1 staining of any intensity [45] [44]. It has received Conformité Européenne in vitro diagnostic (CE-IVD) designation as a companion diagnostic for multiple immunotherapeutic agents including durvalumab, pembrolizumab, cemiplimab-rwlc, and atezolizumab in NSCLC [44].

The SP142 assay employs a distinct scoring system that evaluates PD-L1 expression on both tumor cells (TC) and tumor-infiltrating immune cells (IC) [25] [47]. The IC score represents the percentage of tumor area occupied by PD-L1-stained immune cells [47]. This assay is clinically validated for identifying triple-negative breast cancer (TNBC) patients eligible for atezolizumab therapy using an IC ≥1% cutoff [47].

Comparative Analytical Performance

Multiple studies have investigated the concordance between different PD-L1 assays across various cancer types, with findings demonstrating substantial variability depending on tumor histology and scoring systems.

Concordance in Non-Small Cell Lung Cancer (NSCLC)

In NSCLC, the SP263 and 22C3 assays demonstrate high concordance rates, suggesting potential interchangeability for clinical decision-making.

Table 2: Assay Concordance in NSCLC (IMpower010 Study) [45]

Comparison	Threshold	Concordance Rate	Kappa Statistic	Clinical Outcome Concordance
SP263 vs 22C3	≥1% (Positive)	83%	Not reported	Comparable DFS benefit with atezolizumab
SP263 vs 22C3	≥50% (High)	92%	Not reported	Comparable DFS benefit with atezolizumab

The phase III IMpower010 study demonstrated that the SP263 and 22C3 assays showed high concordance at both the PD-L1-positive (≥1%) and PD-L1-high (≥50%) thresholds in early-stage NSCLC [45]. Importantly, the disease-free survival benefit of adjuvant atezolizumab compared with best supportive care was comparable between assays for both PD-L1-positive and PD-L1-high subgroups, indicating similar predictive value [45].

Concordance in Clear Cell Renal Cell Carcinoma (ccRCC)

A recent comprehensive study evaluating all four FDA-approved PD-L1 assays in clear cell renal cell carcinoma revealed substantial differences in analytical performance.

Table 3: PD-L1 Positivity and Concordance in Clear Cell Renal Cell Carcinoma [25]

Assay Clone	Tumor Cell Positivity	Immune Cell Positivity	Concordance with 28-8 (κ statistic)	Prognostic Significance for CSS
22C3	Very low	14.7%	0.52	Significantly worse
28-8	Very low	16.1%	Reference	Significantly worse
SP142	Very low	2.1%	0.16	Not significant
SP263	Very low	15.0%	0.46	Significantly worse

This study of 286 ccRCC tissue samples demonstrated remarkably low PD-L1 expression in tumor cells across all assays [25]. When assessing immune cell PD-L1 expression, the 28-8 assay showed moderate pairwise concordance with the 22C3 (κ=0.52) and SP263 (κ=0.46) assays, but poor concordance with the SP142 assay (κ=0.16) [25]. Patients with PD-L1 expression in immune cells evaluated using the 22C3, 28-8, and SP263 assays showed significantly worse cancer-specific survival (CSS), establishing prognostic value for these three assays in RCC [25].

Performance in Triple-Negative Breast Cancer (TNBC)

The IMpassion130 study provided insights into assay comparability in TNBC, where the SP142 assay is clinically validated as a companion diagnostic for atezolizumab.

Table 4: Assay Performance in Triple-Negative Breast Cancer (IMpassion130) [47]

Assay	Scoring Method	PD-L1+ Prevalence	Concordance with SP142	Clinical Benefit with Atezolizumab
SP142	IC ≥1%	46.4%	Reference	Significant PFS and OS benefit
SP263	IC ≥1%	74.9%	69.2%	Driven by double-positive cases
22C3	IC ≥1%	73.1%	68.7%	Driven by double-positive cases
22C3	CPS ≥1	80.9%	Not reported	Not significant for single-positive cases

This analysis revealed that the SP263 and 22C3 assays identified substantially more patients as PD-L1-positive (IC ≥1%) compared to the SP142 assay [47]. The analytical concordance between SP142 and the other assays was approximately 69%, indicating suboptimal interchangeability [47]. Importantly, the improved efficacy of atezolizumab plus nab-paclitaxel was primarily driven by patients identified as PD-L1-positive by both SP142 and the alternative assay ("double-positive" cases), rather than those positive only by SP263 or 22C3 ("single-positive" cases) [47].

Experimental Protocols and Methodologies

Standardized experimental protocols are essential for ensuring reproducible and reliable PD-L1 testing across different laboratories and assay platforms.

Tissue Processing and Pre-Analytical Conditions

For surgical pathology specimens, fixation in 10% neutral buffered formalin with fixation times ranging from 3-30 hours (depending on specimen size) is recommended [46]. For cytology specimens, aspirates can be collected directly into methanol-water fixative (CytoLyt), with residual material used to create cell blocks via either plasma-thrombin or Histogel methods before formalin fixation and paraffin embedding [46]. All specimens should be processed on automated tissue processors using standard laboratory programs based on tissue size, with sections cut at 4-micron thickness for staining [46].

Immunohistochemistry Staining Protocols

The 22C3 and 28-8 assays are performed on the Dako Automated Link 48 platform using the manufacturer's specified reagents and protocols [25] [46]. The SP263 and SP142 assays are performed on the Ventana Benchmark platform following the manufacturer's instructions [25] [47]. Proper control samples must be included in each run to ensure staining quality and interpretation accuracy [46].

Scoring Methodology and Pathologist Training

Scoring of PD-L1 expression requires trained pathologists who are proficient in the specific scoring algorithm for each assay [47]. For the 22C3, 28-8, and SP263 assays, scoring focuses primarily on tumor cell membranous staining, while the SP142 assay requires additional assessment of immune cell staining [25] [47]. Pathologists should undergo specific training for each assay and scoring system, with ongoing quality assurance and proficiency testing to maintain consistency [47].

PD-1/PD-L1 Signaling Pathway and Assay Target

The PD-1/PD-L1 axis represents a critical immune checkpoint pathway in cancer biology. PD-L1, encoded by the CD274 gene, is expressed on the surface of tumor cells and tumor-infiltrating immune cells [43] [44]. Interaction between PD-L1 and its receptor PD-1 on T cells leads to inhibition of T cell proliferation, reduced cytokine secretion, and induction of apoptosis in antigen-specific T cells, ultimately resulting in immune escape and tumor progression [43] [44]. The four PD-L1 assays discussed in this guide detect the PD-L1 protein expressed on tumor cells and/or immune cells, enabling identification of patients most likely to respond to ICIs that block this immunosuppressive pathway [44].

PD-L1 Testing Workflow

Essential Research Reagent Solutions

Table 5: Key Research Reagents for PD-L1 Immunohistochemistry

Reagent/Material	Function	Application Notes
Formalin-fixed, paraffin-embedded (FFPE) tissue	Preserves tissue architecture and antigen integrity	Standard 10% neutral buffered formalin; fixation time 3-30 hours depending on specimen size [46]
Cytology cell blocks	Alternative substrate for PD-L1 testing	Prepared from residual cytology material using plasma-thrombin or Histogel method [46]
PD-L1 antibody clones (22C3, 28-8, SP263, SP142)	Specific detection of PD-L1 protein	Each clone has distinct binding epitopes and staining characteristics [25]
Dako Autostainer Link 48	Automated IHC staining platform	Optimized for 22C3 and 28-8 pharmDx assays [45] [46]
Ventana Benchmark series	Automated IHC staining platform	Optimized for SP263 and SP142 assays [45] [47]
Specific detection kits	Signal amplification and visualization	Platform-specific detection systems required for each assay [47]
Control tissues	Quality assurance	Positive and negative controls essential for validating staining quality [46]

The four major FDA-approved PD-L1 assays—22C3, 28-8, SP263, and SP142—demonstrate variable analytical and clinical performance across different cancer types. In NSCLC, the SP263 and 22C3 assays show high concordance and comparable predictive value for ICI benefit, suggesting potential interchangeability in this setting [45]. In contrast, significant disparities exist among all four assays in clear cell renal cell carcinoma, particularly for the SP142 assay, which shows notably lower immune cell positivity and poor concordance with other assays [25]. The SP142 assay remains unique in its emphasis on immune cell staining, which is particularly relevant in specific cancer types such as triple-negative breast cancer [47].

These findings highlight the critical importance of considering assay-specific characteristics when interpreting PD-L1 expression results in both research and clinical settings. The ongoing development of harmonization protocols and artificial intelligence-assisted scoring platforms may help reduce inter-assay variability and improve the accuracy of PD-L1 as a predictive biomarker for immune checkpoint inhibitor therapy [44].

The advent of immune checkpoint inhibitors has established PD-L1 immunohistochemistry (IHC) as a critical predictive biomarker in oncology, making the standardization and reliability of automated staining platforms a cornerstone of modern cancer diagnostics and drug development [48] [49]. Platforms such as the Dako Autostainer Link 48, Ventana BenchMark ULTRA, and Leica BOND-III are integral to performing these complex assays. However, the comparative performance of these systems, influenced by their unique chemistries, protocols, and sensitivities, directly impacts the accuracy of patient selection for therapy. This guide objectively compares these leading platforms, framing the analysis within the broader thesis of PD-L1 assay standardization and providing researchers with the experimental data necessary to inform their analytical and clinical decisions.

The Dako Autostainer Link 48 (Agilent), Ventana BenchMark ULTRA (Roche), and Leica BOND-III represent the leading technologies in automated IHC and ISH (in situ hybridization) staining. Each system employs a distinct approach to automation, reagent management, and the staining process itself, which contributes to its unique performance profile.

The table below summarizes the core specifications and technological approaches of each platform:

Table 1: Key Specifications of Automated Staining Platforms

Feature	Dako Autostainer Link 48	Ventana BenchMark ULTRA	Leica BOND-III
Manufacturer	Agilent Technologies	Roche Ventana	Leica Biosystems
Staining Principle	Capillary gap (Coverplate) [50]	Puddle (Liquid Coverslip) [51]	Puddle (Covertile) [52]
Typical Slide Capacity	48 slides [53]	30 slides [51]	30 slides [52]
Workflow	Batch-based	Single-piece, continuous access [51]	Batch-based with 3 independent trays [52]
Key Technology	Semi-automated (separate antigen retrieval) [54]	Individually controlled slide heater pads [51]	Patented Covertile system for low reagent volume [52]
Reagent System	Open	Largely closed with bar-coded testpacks [50]	Open with real-time level alerts [52]
Assay Menu Flexibility	High	Broad (250+ ready-to-use assays) [51]	High, with Novocastra reagents [52]

Comparative Performance in Biomarker Assays

The analytical performance of these platforms is paramount, particularly for standardized companion diagnostics. Studies have directly compared their output for key biomarkers like PD-L1 and Ki-67, revealing significant differences in assay sensitivity and inter-instrument concordance.

PD-L1 Assay Sensitivity and Harmonization

PD-L1 IHC is a primary diagnostic for immunotherapy, but multiple FDA-approved assays with different antibodies exist, leading to challenges in harmonization. Research shows that the analytical sensitivity of these assays varies substantially by platform.

A multi-institutional study using a standardized PD-L1 Index TMA and quantitative digital analysis found that FDA-approved assays could be grouped by analytic sensitivity. The Ventana SP263 assay was found to be the most sensitive, followed by the Agilent 22C3 and 28-8 assays, while the Ventana SP142 assay was analytically ten times less sensitive than the SP263 assay [48]. This lower sensitivity of the SP142 assay was confirmed in another study, which noted it failed to detect low levels of PD-L1 in cell lines that were distinguished by other assays [49]. Critically, the assays for 22C3, 28-8, SP263, and a laboratory-developed test using E1L3N were highly similar and consistent across multiple laboratory sites for a given platform [49].

Ki-67 Assay Concordance and Platform Variability

Ki-67 is a proliferation marker with well-documented inter-laboratory heterogeneity. A comparative study of Ki-67 IHC laboratory-developed tests (LDTs) on different platforms highlighted significant variability.

Table 2: Analytical Comparison of Ki-67 IHC Laboratory-Developed Tests

Platform & Antibody Clone	Sensitivity at 20% Cutoff (%) [55]	Specificity at 20% Cutoff (%) [55]	Key Finding
Dako Autostainer Link 48 (MIB-1)	24.8	99.5	High specificity but low sensitivity vs. reference assay.
Leica BOND-III (K2)	25.1	100.0	Performance nearly identical to Dako AS48 LDT.
Ventana BenchMark ULTRA (30-9)	99.3	53.6	High sensitivity but markedly lower specificity.

This data demonstrates that the choice of platform and antibody clone combination can drastically alter the classification of samples, as seen with the Ventana 30-9 clone, which showed high sensitivity but low specificity compared to the reference test [55].

Furthermore, a specific study comparing the FDA-approved Ki-67 IHC MIB-1 pharmDx assay on the Dako Omnis versus its reagents used with an optimized protocol on the more widely available Dako Autostainer Link 48 (AS48) showed that high concordance (90.3% overall agreement) is achievable between instruments from the same manufacturer [53]. This suggests that reagent and protocol optimization are as critical as the choice of instrument.

Specialized Staining Patterns: The Case of MIB-1 in HTT

For some diagnostics, the staining pattern is as important as the intensity. In diagnosing Hyalinising Trabecular Tumour (HTT) of the thyroid, a specific cell membrane-positive reaction for MIB-1 (Ki-67) is a crucial criterion. A 2024 study investigated the ability of different automated platforms to replicate this pattern, which is routinely achieved with manual staining.

Table 3: Performance of Automated Platforms for MIB-1 Membrane Staining in HTT

Platform	Optimal Conditions	Staining Outcome for Membrane Pattern
Dako Autostainer Link 48	Antigen retrieval with pH 9.0 at room temperature (RT)	Most stable and strongest membrane staining [54].
Ventana BenchMark ULTRA	CC1 (pH 8.5) retrieval; primary antibody incubation at RT	Significantly stronger membrane staining at RT than at 37°C [54].
Leica BOND-III	ER1 (pH 6.0) retrieval at RT	Weak-to-moderate membrane staining; weaker with pH 9.0 [54].
Dako Omnis	Antigen retrieval at pH 9.0; incubation at 32°C	Weak-to-moderate positive membrane staining [54].

This study concluded that the Dako Autostainer Link 48 was the most stable platform for this particular application, closely mimicking manual staining conditions. It also highlighted that slight adjustments in protocol parameters, such as antigen retrieval pH and incubation temperature, are critical for success on automated systems and are not universally optimal across platforms [54].

Experimental Protocols for Platform Comparison

To ensure objectivity, the data cited in this guide are derived from rigorous, published experimental methodologies. The key protocols are summarized below for researcher reference.

Multi-Institutional PD-L1 Assay Comparison

This study quantified inter-assay and inter-laboratory variation using a standardized Index Tissue Microarray (TMA) [49].

• TMA Construction: A TMA was constructed using 10 isogenic cell lines expressing a dynamic range of PD-L1, formalin-fixed and paraffin-embedded (FFPE) in triplicate [49].
• Staining Protocol: Five PD-L1 IHC assays (FDA-approved and LDTs) were tested. For the multi-institutional comparison, 12 sections of the Index TMA were sent to 12 institutions. Each site stained two slides weekly for six consecutive weeks using their standard clinical PD-L1 assay and platform (e.g., Dako Autostainer Link 48 for 22C3/28-8, Ventana BenchMark ULTRA for SP263/SP142) [49].
• Quantitative Analysis: Stained slides were digitized. PD-L1 expression was quantified using digital image analysis (QuPath software), measuring the optical density of chromogenic staining per mm² [49].
• Data Analysis: Agreement between assays and laboratories was assessed using linear regression coefficients (R²) and Levey-Jennings plots to evaluate consistency over time [49].

Ki-67 Concordance Study Between Dako Omnis and AS48

This study measured the concordance of the FDA-approved Ki-67 IHC MIB-1 pharmDx assay across two Dako instruments [53].

• Tissue Samples: 40 FFPE breast carcinoma samples were selected, with 19 samples having Ki-67 scores near the 10-30% clinical cutoff [53].
• Staining Protocol: Sections were stained using the Ki-67 IHC MIB-1 pharmDx kit. The assay was run on the Dako Omnis per FDA instructions and on the Dako Autostainer Link 48 (AS48) using an optimized protocol that adjusted the order, number, and duration of wash cycles [53].
• Blinded Pathologist Assessment: Three certified pathologists scored the whole-slide images at three different timepoints, blinded to the instrument and specimen, using the approved Ki-67 scoring approach [53].
• Statistical Analysis: Overall agreement and a concordance correlation coefficient (CCC) for continuous scores were calculated. A predefined cutoff of ≥20% was used to determine positive/negative status [53].

Figure 1: Experimental Workflow for Platform Comparison Studies. This diagram outlines the core methodologies used in the key studies cited to generate comparable data on staining platform performance.

The Scientist's Toolkit: Key Reagents and Materials

The performance of an automated staining system is dependent on an integrated set of reagents and materials. The following table details essential components referenced in the featured studies.

Table 4: Essential Research Reagents and Materials for Automated IHC

Item	Function	Example in Context
Index Tissue Microarray (TMA)	A standardized slide containing multiple tissue or cell line cores for simultaneous staining, enabling inter-laboratory and inter-assay comparison [49].	Used with isogenic cell lines expressing a PD-L1 dynamic range to objectively compare assay sensitivity [49].
FDA-Approved IHC Assay Kits	Complete reagent sets (primary antibody, detection system) validated for a specific diagnostic purpose on a designated platform.	PD-L1 IHC 22C3 PharmDx (for Dako platforms) [49]; PD-L1 IHC SP263 Assay (for Ventana platforms) [48] [49].
Validated Antibody Clones	The specific monoclonal antibody that binds the target epitope. Different clones can have varying performance.	Ki-67 clones MIB-1, K2, and 30-9 show different sensitivity/specificity profiles on Dako, Leica, and Ventana platforms, respectively [55].
Epitope Retrieval Solutions	Buffered solutions used to reverse formaldehyde cross-linking and expose hidden antigenic epitopes. pH is critical.	Dako Target Retrieval Solution (pH 6.0 or 9.0) [54]; Ventana Cell Conditioning Solution (CC1, pH ~8.5) [54].
Detection System	A series of reagents that generate a visible signal (chromogenic or fluorescent) at the site of antibody binding.	Bond Polymer Refine Detection (Leica) [54]; EnVision FLEX (Dako) [53] [54]; UltraView DAB (Ventana) [54].
NIST-Traceable Calibrators	Synthetic calibrators (e.g., peptide-coated microbeads) with a known number of molecules/bead, used to standardize assay sensitivity and reproducibility [48].	Tool for standardizing the biochemical aspect of PD-L1 IHC assays, ensuring week-to-week reproducibility of stain intensity [48].
4-Methoxy-2,3,6-trimethylbenzyl bromide	4-Methoxy-2,3,6-trimethylbenzyl bromide, CAS:69877-88-9, MF:C11H15BrO, MW:243.14 g/mol	Chemical Reagent
3-O-Demethylmonensin B	3-O-Demethylmonensin B|For Research|RUO	3-O-Demethylmonensin B is a monensin derivative isolated from Streptomyces cinnamonensis. For Research Use Only. Not for human or veterinary use.

The choice between the Dako Autostainer Link 48, Ventana BenchMark ULTRA, and Leica BOND-III systems involves critical trade-offs. The data indicates that no single platform is universally superior; instead, the optimal instrument depends on the specific application and required context.

For standardized companion diagnostics like PD-L1, the platform is often predetermined by the FDA-approved assay. However, researchers must be aware of the inherent sensitivity differences between assays (e.g., SP263 vs. SP142) [48] [49]. For laboratory-developed tests (LDTs) like Ki-67, the platform and antibody clone selection will profoundly impact the results, as evidenced by the significant variability in sensitivity and specificity [55]. Furthermore, for highly specialized staining patterns like MIB-1 membrane staining in HTT, platform-specific protocol optimization is not just beneficial but essential, with some systems offering more stable performance than others [54].

Therefore, the path to reliable and reproducible IHC data lies in understanding the technical nuances of these automated systems, rigorously validating each assay on the chosen platform, and implementing standardization tools like index TMAs and NIST-traceable calibrators to ensure analytical precision across experiments and laboratories [48] [49].

The advent of immune checkpoint inhibitors (ICIs) targeting the PD-1/PD-L1 axis has revolutionized cancer treatment, making accurate assessment of PD-L1 expression a critical component of companion diagnostics. As of 2025, the U.S. Food and Drug Administration (FDA) has approved 12 PD-L1 companion diagnostics for immunotherapies, each utilizing different scoring methods and thresholds [56]. Two scoring systems have emerged as fundamental to this evaluation: the Tumor Proportion Score (TPS) and the Combined Positive Score (CPS). These quantitative immunohistochemistry (IHC) scoring methods guide therapeutic decisions across multiple cancer types, including non-small cell lung cancer (NSCLC), gastric cancer, and head and neck squamous cell carcinoma (HNSCC). The selection between TPS and CPS has significant implications for patient selection, as it directly influences eligibility for specific ICIs. This guide provides a comprehensive comparison of these systems, examining their technical specifications, clinical applications, performance characteristics, and implementation protocols to support researchers and drug development professionals in optimizing PD-L1 detection strategies.

Defining the Scoring Systems: TPS vs. CPS

Core Definitions and Calculation Methods

Tumor Proportion Score (TPS) is defined as the percentage of viable tumor cells exhibiting partial or complete PD-L1 membrane staining relative to all viable tumor cells in the sample [57] [56]. The calculation excludes immune cells and stromal elements, focusing exclusively on neoplastic cells. The formula is expressed as:

TPS = (Number of PD-L1 positive tumor cells ÷ Total number of viable tumor cells) × 100

Combined Positive Score (CPS) represents a more comprehensive metric that quantifies PD-L1 expression across both tumor and immune compartments [57] [58] [56]. CPS is calculated as the number of PD-L1-positive cells (tumor cells, lymphocytes, and macrophages) divided by the total number of viable tumor cells, multiplied by 100:

CPS = (Number of PD-L1 positive cells [tumor cells, lymphocytes, macrophages] ÷ Total number of viable tumor cells) × 100

Table 1: Fundamental Characteristics of TPS and CPS Scoring Systems

Characteristic	Tumor Proportion Score (TPS)	Combined Positive Score (CPS)
Cells assessed	Tumor cells only	Tumor cells, lymphocytes, and macrophages
Scoring range	0-100%	0-100 (theoretically unlimited but typically reported up to 100)
Key components	PD-L1+ tumor cells, total viable tumor cells	PD-L1+ tumor cells, PD-L1+ immune cells, total viable tumor cells
Excluded elements	Immune cells, stromal cells, necrotic areas	Necrotic areas, non-viable tumor cells
Primary clinical context	NSCLC, first-line pembrolizumab monotherapy	Gastric cancer, HNSCC, urothelial carcinoma

Clinical Cut-offs and Therapeutic Thresholds

Both scoring systems employ specific thresholds that trigger therapeutic implications across different cancer types. For TPS, the most significant cut-point is ≥50% for first-line pembrolizumab monotherapy in metastatic NSCLC, while ≥1% may indicate benefit in other contexts [41] [56]. For CPS, multiple thresholds exist across indications: ≥1 for gastric cancer (pembrolizumab), ≥5 for gastric cancer (nivolumab in CheckMate-649), and ≥10 for esophageal cancer (pembrolizumab in KEYNOTE-590) [59] [60]. The RATIONALE-305 trial introduced yet another metric called Tumor Area Positivity (TAP), which quantifies both tumor and immune cell staining, with ≥5% defining positivity for tislelizumab benefit in gastric cancer [60]. This proliferation of scoring systems and thresholds underscores the importance of assay-specific biomarker validation in immunotherapy trials.

Comparative Performance Across Cancer Types

Analytical Concordance and Inter-assay Variability

The comparability of different PD-L1 assays has been extensively studied to determine potential interchangeability in clinical practice. A comprehensive comparability study of immunohistochemical assays for PD-L1 detection in hepatocellular carcinoma demonstrated that the 22C3, 28-8, and SP263 assays exhibited comparable sensitivity in detecting PD-L1 expression, whereas the SP142 assay was consistently the least sensitive across both TPS and CPS evaluations [17]. The inter-assay agreement, measured by intraclass correlation coefficients (ICC), was 0.646 for TPS and 0.780 for CPS, indicating superior concordance for the combined scoring system [17]. This enhanced agreement with CPS likely stems from its incorporation of multiple cell types, which may mitigate tumor heterogeneity effects and staining interpretation variability.

The inter-rater reliability also differs between scoring systems. In the hepatocellular carcinoma study, the overall ICC among five pathologists was 0.946 for TPS and 0.809 for CPS, suggesting that pathologists demonstrate greater consistency when evaluating tumor cells alone compared to the more complex assessment required for CPS [17]. This reliability gap highlights the challenging nature of immune cell quantification in the tumor microenvironment, particularly with certain assays like SP142, where pathologists were less reliable in scoring CPS compared to TPS [17]. Importantly, up to 18% of samples were misclassified by individual pathologists compared to consensus scoring at the CPS ≥1 cutoff, emphasizing the clinical impact of this variability [17].

Predictive Power Across Tumor Types

The relative clinical utility of TPS versus CPS varies significantly across cancer types, reflecting fundamental differences in tumor biology and immune microenvironment composition.

In gastric cancer, CPS has emerged as the dominant biomarker across multiple pivotal trials. The KEYNOTE-859 trial established CPS ≥1 as the threshold for pembrolizumab approval in advanced gastric cancer, demonstrating significant survival benefits (median OS: 13.0 months vs 11.4 months; HR=0.74) [60]. This advantage was more pronounced in the CPS ≥10 subgroup (median OS 15.7 months vs 11.8 months; HR=0.65) [60]. Similarly, the CheckMate-649 trial led to nivolumab approval in HER2-negative advanced gastric cancer with CPS ≥5 [60]. The integration of HER2 status with PD-L1 scoring is particularly relevant in gastric cancer, as the KEYNOTE-811 trial demonstrated that adding pembrolizumab to trastuzumab and chemotherapy significantly improved objective response rates (74.4% vs 51.9%) in HER2-positive gastric cancer, with superior overall survival in PD-L1 CPS ≥1 patients (20.1 months vs 15.7 months; HR=0.79) [60].

In NSCLC, TPS remains the established biomarker in many contexts, particularly for pembrolizumab monotherapy in patients with TPS ≥50% [41] [56]. However, emerging quantitative approaches are revealing new dimensions of PD-L1 assessment. The PD-L1 Quantitative Continuous Scoring (QCS) system identifies NSCLC patients more likely to benefit from durvalumab by capturing the percentage of tumor cells with medium to strong staining intensity [41]. This continuous scoring method demonstrated a hazard ratio of 0.62 (CI 0.46-0.82) with a biomarker-positive prevalence of 54.3%, compared to visual TPS scoring which resulted in a hazard ratio of 0.69 (CI 0.46-1.02) with a 29.7% prevalence [41].

In HNSCC, both scoring systems are utilized, with CPS ≥1 determining first-line pembrolizumab eligibility based on the KEYNOTE-689 trial, while TPS ≥50% guides second-line treatment [57]. A study examining PD-L1 expression across different specimen types in HNSCC found significant discrepancies in both CPS and TPS between biopsy and surgical resection specimens (p<0.01), as well as between resection and metastatic lymph nodes (p<0.01) [57]. This heterogeneity underscores the importance of standardized specimen selection for PD-L1 assessment regardless of scoring system.

Table 2: Clinical Applications of TPS and CPS Across Cancer Types

Cancer Type	Preferred Scoring System	Key Therapeutic Thresholds	Supporting Clinical Trials
Non-small cell lung cancer (NSCLC)	TPS (primary)	TPS ≥50% (first-line pembrolizumab monotherapy)	KEYNOTE-024, MYSTIC
Gastric cancer	CPS	CPS ≥1 (pembrolizumab), CPS ≥5 (nivolumab)	KEYNOTE-859, CheckMate-649, KEYNOTE-811
Head and neck squamous cell carcinoma (HNSCC)	Both	CPS ≥1 (first-line), TPS ≥50% (second-line)	KEYNOTE-689, CHECKMATE-141
Esophageal cancer	CPS	CPS ≥10 (pembrolizumab)	KEYNOTE-590
Angiosarcoma	TPS	TPS ≥1% (emerging biomarker)	Investigational

Experimental Protocols and Methodologies

Standardized Staining and Assessment Protocols

Robust PD-L1 scoring requires strict adherence to standardized experimental protocols across pre-analytical, analytical, and post-analytical phases. For the widely used PD-L1 IHC 22C3 pharmDx assay, tissue sections are cut at 4μm thickness from formalin-fixed paraffin-embedded (FFPE) blocks and processed on automated staining platforms such as the Ventana BenchMark ULTRA or Dako Autostainer Link 48 [57] [58]. Appropriate positive and negative controls must be included in each run, with multi-tissue blocks containing tonsil and placenta tissues often serving as positive controls [56]. The entire process for 204 tissue sections from a HNSCC study was automated to ensure consistency [57].

For novel assay development, such as the PD-L1 CAL10 assay (Leica Biosystems) currently in development, feasibility studies compare performance to established assays like the SP263 (Ventana) on the Benchmark Ultra staining system [56]. These comparability studies require careful attention to inclusion criteria—typically encompassing both resection and biopsy specimens from relevant cancer types (e.g., 60-70% adenocarcinomas and 30-40% squamous cell carcinomas for NSCLC) with representation of both primary and metastatic sites [56]. Such standardization enables reliable assessment of diagnostic concordance, with the CAL10 assay demonstrating a lower bound of the 95% CI of overall percent agreement of 86.2% at ≥50% TPS cutoff and 94.0% at ≥1% TPS cutoff compared to SP263 [56].

Digital Pathology and AI-Assisted Scoring Workflows

Advanced computational approaches are addressing the challenges of manual PD-L1 assessment through automated scoring pipelines. A typical AI-based workflow for CPS quantification in gastric cancer incorporates multiple sequential deep learning models [58]:

• Tissue localization using Otsu thresholding on grayscale-converted whole slide images (WSIs) at low magnification (0.625×)
• Patch-level classification with MobileNet-v2 to identify tumor-containing regions
• Pixel-level segmentation using U-Net to delineate tumor versus non-tumor regions
• Cell detection with YOLO-based models to identify PD-L1+ tumor cells, PD-L1− tumor cells, and PD-L1+ immune cells
• Automated CPS calculation based on cellular counts from detection outputs

This integrated pipeline demonstrated strong concordance with expert pathologists' consensus in internal validation (Cohen's kappa = 0.782) and maintained robust performance in external cohorts (Cohen's kappa = 0.737) [58].

For rare cancers like angiosarcoma, where training data is limited, specialized pipelines such as PEERCE leverage pre-trained generalist models and fine-tuning approaches to achieve strong TPS prediction performance (correlation coefficients of 0.83-0.93 with pathologist assessment) despite limited annotated data [61]. In this context, AI assistance serves as a valuable "second opinion," with pathologists updating their TPS scores in cases of strong disagreement, thereby improving diagnostic accuracy [61].

Digital vs. Traditional Pathology Workflows for PD-L1 Scoring

The Scientist's Toolkit: Essential Research Reagents and Platforms

Table 3: Essential Research Reagents and Platforms for PD-L1 Scoring

Category	Specific Products/Platforms	Research Application	Key Characteristics
IHC Assays	PD-L1 IHC 22C3 pharmDx (Agilent/Dako)	Gold standard for TPS/CPS assessment	FDA-approved companion diagnostic; used with Autostainer Link platforms
	PD-L1 IHC SP263 (Ventana)	Comparative assay validation	Comparable sensitivity to 22C3; used on Benchmark Ultra system
	PD-L1 CAL10 (Leica Biosystems)	Novel assay development	Demonstrates 86.2% OPA with SP263 at ≥50% TPS; BOND-III platform
Staining Platforms	Ventana BenchMark ULTRA	Automated IHC staining	Standardized processing for SP263 and SP142 assays
	Dako Autostainer Link 48	Automated IHC staining	Optimized for 22C3 pharmDx assay
	Leica BOND-III	Automated IHC staining	Development platform for CAL10 assay
Digital Pathology	Philips Intellisite Pathology Solution	Whole slide imaging	High-resolution TIFF file export for bioimage analysis
	3DHISTECH PANNORAMIC 1000	Whole slide imaging	40× magnification scanning (0.25μm/pixel) for AI analysis
	Aperio GT 450	Whole slide imaging	Digital read concordance with glass slides (94% at ≥1% TPS)
Bioimage Analysis	QuPath	Open-source bioimage analysis	Manual annotation of tumor cells, immune cells, PD-L1+ populations
	PEERCE Library	AI-assisted TPS prediction	Angiosarcoma-focused; open-source pipeline for rare cancers
	Custom AI Pipelines (MobileNet-v2, U-Net, YOLO)	Automated CPS quantification	Integrated patch classification, segmentation, and cell detection
Pirlimycin	Pirlimycin, CAS:78822-40-9, MF:C17H31ClN2O5S, MW:411.0 g/mol	Chemical Reagent	Bench Chemicals
Zapnometinib	Zapnometinib, CAS:303175-44-2, MF:C13H7ClF2INO2, MW:409.55 g/mol	Chemical Reagent	Bench Chemicals

The comparative analysis of TPS and CPS scoring systems reveals a complex landscape where clinical utility is highly context-dependent, varying by cancer type, therapeutic agent, and specific clinical setting. While TPS offers simplicity and superior inter-rater reliability, particularly in NSCLC, CPS provides a more comprehensive assessment of the tumor immune microenvironment that has proven valuable in gastrointestinal cancers and HNSCC. The ongoing development of quantitative continuous scoring systems and AI-assisted pipelines promises to address current limitations in inter-observer variability and specimen-related heterogeneity [41] [58]. Furthermore, the emergence of novel assessment metrics like Tumor Area Positivity (TAP) suggests that the evolution of PD-L1 scoring is ongoing, with future systems potentially incorporating spatial relationships and multiplexed biomarker information [60]. For researchers and drug development professionals, selection between TPS and CPS must be guided by specific cancer indications, available tissue specimens, and the growing arsenal of computational tools that enhance scoring precision and reproducibility.

Laboratory-developed tests are in vitro diagnostic products designed, manufactured, and used within a single clinical laboratory [62]. Historically, the U.S. Food and Drug Administration exercised enforcement discretion over LDTs, but the regulatory landscape has undergone significant changes. In 2024, the FDA announced a final rule to phase out its general enforcement discretion approach, though this rule was subsequently vacated by a federal court in 2025 [63] [64]. This regulatory uncertainty forms the critical backdrop against which laboratories must develop and implement LDTs, particularly for complex applications such as PD-L1 immunohistochemistry testing in cancer.

The significance of LDTs is particularly pronounced in specialized domains like PD-L1 detection for immunotherapy selection. With different immune checkpoint inhibitors linked to specific companion diagnostic assays, laboratories face practical challenges in offering multiple commercial tests [65]. LDTs provide a vital pathway for laboratories to expand testing capabilities using existing platforms, thereby increasing patient access to essential predictive biomarkers without being constrained by proprietary instrument systems [65] [66].

LDTs Versus Commercial Assays: A Comparative Analysis

Regulatory and Operational Distinctions

Table 1: Key Differences Between LDTs and Commercial IVDs

Aspect	Laboratory-Developed Tests (LDTs)	Commercial IVDs
Regulatory Oversight	CLIA certification, CAP inspections [67]	FDA premarket review (510(k), PMA) [68] [62]
Development Flexibility	Rapid adaptation and modification capabilities [66]	Fixed design without modifications allowed [6]
Content Control	Laboratory controls target selection and relevance [66]	Manufacturer determines content
Implementation Timeline	Relatively quick development and validation [66]	Lengthy development and regulatory review
Cost Structure	Lower cost per test [66]	Higher development costs incorporated into pricing
Technical Support	Laboratory self-sufficient	Manufacturer-provided technical support [66]
Test Consolidation	Multiple analytes possible in single test [66]	Typically focused on specific analytes
Distribution Scope	Limited to developing laboratory	Broad distribution across multiple laboratories [66]

Advantages and Limitations in Practice

LDTs offer several distinct advantages that make them particularly valuable in specialized clinical and research settings. They provide laboratories with direct control over test content, enabling the selection of specific and relevant targets tailored to patient populations [66]. This flexibility extends to the ability to rapidly develop and modify tests in response to emerging clinical needs, which proved crucial during public health emergencies such as the COVID-19 and mpox outbreaks [67]. Additionally, LDTs enable test consolidation, allowing multiple analytes to be measured in a single test, which can provide more comprehensive data per sample and potentially accelerate diagnostic processes [66].

Commercial IVDs, in contrast, benefit from established quality systems with design and manufacturing controls required for FDA clearance [66]. They offer clinical validity demonstrated through extensive validation studies, and users have access to manufacturer technical support for troubleshooting [66]. The broad distribution of commercial tests across multiple laboratories generates substantial collective data that can reinforce confidence in test performance [66].

Quantitative Performance Comparison of PD-L1 Assays

Analytical Concordance Between Testing Platforms

Table 2: Performance Comparison of PD-L1 Immunohistochemistry Assays

Assay Type	Comparison	TPS ≥1% Cutoff	TPS ≥50% Cutoff	Study Details
LDT (22C3 on VENTANA)	vs. 22C3 DAKO	94.6% OPA [65]	91.8% OPA [65]	85 NSCLC cases [65]
LDT (22C3 on VENTANA)	vs. SP263 VENTANA	95.0% OPA [65]	93.8% OPA [65]	85 NSCLC cases [65]
Commercial (22C3 DAKO)	vs. SP263 VENTANA	91.8% OPA [65]	96.5% OPA [65]	85 NSCLC cases [65]
Novel CAL10 (Under Development)	vs. SP263 VENTANA	≥94.0% OPA (95% CI) [56]	≥86.2% OPA (95% CI) [56]	136 NSCLC samples [56]

Meta-Analysis Evidence on Assay Interchangeability

A comprehensive meta-analysis of 22 publications encompassing 376 assay comparisons revealed crucial insights about PD-L1 assay interchangeability [6]. The analysis established that for a testing laboratory unable to use an FDA-approved companion diagnostic, developing a properly validated LDT for the same purpose as the original PD-L1 FDA-approved immunohistochemistry companion diagnostic is preferable to replacing it with another FDA-approved companion diagnostic developed for a different purpose [6].

This research further determined that LDTs can achieve diagnostic accuracy meeting the clinically acceptable threshold of ≥90% sensitivity and specificity for stated clinical applications when properly validated [6]. However, the performance of LDTs shows greater variability compared to FDA-approved assays due to differences in immunohistochemistry protocol conditions across laboratories, even when using the same primary antibody and automated instrument platform [6].

Experimental Protocols for PD-L1 Assay Validation

Cross-Platform Assay Development Methodology

The development and validation of PD-L1 LDTs follows rigorous experimental protocols to ensure analytical reliability. In a study comparing laboratory-developed and commercial PD-L1 assays, researchers implemented a systematic approach [65]:

Tissue Sample Selection: The study utilized 85 non-small cell lung carcinoma cases from surgical resections, with patient ages ranging from 40 to 88 years. The cohort included both adenocarcinoma and squamous cell carcinoma subtypes to represent the spectrum of NSCLC [65].

Assay Configuration: The LDT was developed using 22C3 antibody on the VENTANA BenchMark ULTRA platform, contrasting with the commercial 22C3 pharmDx Assay designed for the Dako Autostainer Link 48 platform. This cross-platform application required careful optimization of staining conditions [65].

Staining and Evaluation: Triplicate glass slides were stained for each case, including the target stain, H&E for morphological reference, and appropriate negative control isotypes. Staining intensity was quantitatively assessed, with the 22C3 Dako assay producing more intense membrane staining compared to both Ventana platform assays [65].

Statistical Analysis: Overall percent agreement was calculated for key clinical cutoffs (TPS ≥1% and ≥50%), with concordance determined through direct comparison of paired sample results across platforms [65].

Novel Assay Development and Digital Pathology Integration

A more recent development study of the novel PD-L1 CAL10 assay on the BOND-III platform demonstrates advanced validation methodologies [56]:

Study Design: The feasibility analysis included 136 formalin-fixed paraffin-embedded NSCLC tissue samples, with case selection following strict inclusion criteria requiring 60-70% adenocarcinomas, 30-40% squamous cell carcinomas, and representation of both primary and metastatic sites [56].

Pre-screening Protocol: All cases were pre-characterized using the BOND RTU PD-L1 (73-10) clone to establish baseline PD-L1 expression across the 0-100% TPS range [56].

Digital Pathology Integration: After traditional manual assessment, CAL10-stained glass slides were scanned using the Aperio GT 450 scanner to generate whole slide images. Pathologists re-evaluated the digital images after a 4-month washout period to assess concordance between manual and digital reading modalities [56].

Statistical Framework: A one-sided, exact non-inferiority test for a single proportion with a 0.05 type 1 error rate was applied to demonstrate non-inferiority of the CAL10 assay to the SP263 comparator [56].

Diagram 1: PD-1/PD-L1 Signaling Pathway and Immunotherapy Mechanism. This diagram illustrates how tumor cells expressing PD-L1 interact with PD-1 receptors on T-cells to suppress immune response, and how immune checkpoint inhibitors block this interaction to restore anti-tumor immunity.

The Scientist's Toolkit: Essential Research Reagents and Platforms

Table 3: Key Research Reagent Solutions for PD-L1 Assay Development

Reagent/Platform	Function	Application Notes
Primary Antibodies (22C3, SP263, CAL10)	Specific PD-L1 epitope binding	Clone selection affects staining intensity and interpretation [65] [56]
Automated Staining Platforms (Dako Link 48, VENTANA BenchMark, BOND-III)	Standardized assay execution	Platform choice affects staining patterns; cross-platform development requires optimization [65] [56]
Detection Systems	Signal amplification and visualization	Platform-specific detection chemistry impacts sensitivity [6]
Tissue Controls (Tonsil, Placenta)	Assay performance verification	Multi-tissue blocks used for process control [56]
Digital Pathology Scanners (Aperio GT 450)	Whole slide imaging for analysis	Enables digital read concordance studies [56]
NIST Standard Reference Material 1934	Metrological traceability	Enables quantitative comparison across different PD-L1 assays [69]
Glyceryl 1-monooctanoate	Glyceryl 1-monooctanoate, CAS:26402-26-6, MF:C11H22O4, MW:218.29 g/mol	Chemical Reagent
Monomethyl kolavate	Monomethyl Kolavate|TbGAPDH Inhibitor	Monomethyl kolavate is a potent TbGAPDH inhibitor (IC50 = 2 µM) for trypanosomiasis research. For Research Use Only. Not for human or veterinary use.

Diagram 2: LDT Development and Implementation Workflow. This flowchart outlines the key phases in developing and implementing laboratory-developed tests, from initial concept through post-market monitoring, highlighting critical validation and verification steps.

Regulatory Considerations and Future Directions

The regulatory environment for LDTs remains dynamic and complex. The FDA's 2024 final rule sought to establish a four-year phaseout of enforcement discretion, citing concerns about modern LDTs being used more widely for critical healthcare decisions despite variable performance [62]. However, the 2025 federal court decision vacating this rule affirmed that LDTs constitute services rather than devices, placing them outside FDA medical device authorities [63] [64].

This regulatory uncertainty necessitates strategic planning for laboratories developing PD-L1 LDTs. The fundamental framework should include proper validation following Clinical and Laboratory Standards Institute protocols, ongoing performance monitoring, and rigorous adherence to CLIA requirements [68]. Furthermore, laboratories should implement comprehensive quality management systems that address pre-analytical, analytical, and post-analytical phases of testing [67].

Future developments in PD-L1 testing will likely focus on improved standardization through reference materials traceable to NIST standards, which enable quantitative comparison across different PD-L1 IHC assays [69]. Additionally, the integration of digital pathology and artificial intelligence for scoring may reduce inter-pathologist variability, which has been identified as a significant factor in PD-L1 assessment [65] [6].

The development and implementation of LDTs for PD-L1 detection represents a critical capability for modern clinical laboratories, particularly in the context of evolving regulatory frameworks and the need for accessible cancer immunotherapy biomarkers. When properly validated, LDTs demonstrate performance comparable to commercial assays, with the meta-analysis evidence indicating that properly validated LDTs can achieve the clinically acceptable threshold of ≥90% sensitivity and specificity [6]. The strategic implementation of PD-L1 LDTs requires careful attention to analytical validation, clinical concordance studies, and ongoing quality management to ensure reliable patient results for immunotherapy selection.

The accurate assessment of programmed death-ligand 1 (PD-L1) expression through immunohistochemistry (IHC) is a critical predictive biomarker for patient selection in immune checkpoint inhibitor therapy. The reliability of this biomarker, however, is profoundly influenced by pre-analytical variables, particularly the type of tissue specimen analyzed. In clinical practice and research, pathologists and researchers encounter diverse specimen types including surgical resections, biopsies, and cytology cell blocks, each with distinct structural properties and technical challenges. Understanding how PD-L1 expression varies across these different specimen types, and the implications for assay performance, is essential for accurate treatment stratification in oncology, particularly for non-small cell lung cancer (NSCLC) and head and neck squamous cell carcinoma (HNSCC). This guide objectively compares PD-L1 testing performance across different tissue specimens, supported by experimental data and detailed methodologies from recent studies.

Performance Comparison Across Specimen Types

Quantitative Comparison of PD-L1 Expression

Comparative studies have demonstrated significant heterogeneity in PD-L1 expression when measured across different specimen types from the same patients. This variability presents substantial challenges for consistent biomarker interpretation and patient selection.

Table 1: Comparison of PD-L1 Expression Across Specimen Types in HNSCC

Specimen Type	TPS Comparison	CPS Comparison	Statistical Significance	Study Details
Preoperative Biopsy	Lower than resection	Lower than resection	p < 0.01 for both TPS and CPS	68 HNSCC cases; 22C3 assay [57]
Surgical Resection	Reference standard	Reference standard	Reference for comparisons	Digital analysis with QuPath [57]
Metastatic Lymph Node	Lower than resection	Lower than resection	p < 0.01 for both TPS and CPS	Same patients, triple sampling [57]
Biopsy vs. Lymph Node	No significant difference	No significant difference	Not statistically significant	Despite different origins [57]

The observed discrepancies highlight the impact of tumor heterogeneity and sample representation on PD-L1 assessment. Surgical resections provide more comprehensive tumor sampling, potentially capturing the full spectrum of PD-L1 expression patterns, while biopsies and metastatic deposits may only represent subsets of the tumor biology [57]. This has direct implications for clinical trial design and diagnostic accuracy, as specimen type may influence patient eligibility for immunotherapy.

Impact on Clinical Decision-Making

The variability in PD-L1 expression across different specimen types can directly impact treatment decisions, particularly when using standardized scoring cutoffs.

Table 2: Impact of Specimen Type on PD-L1 Scoring and Clinical Implications

Factor	Impact on PD-L1 Scoring	Potential Clinical Consequence
Tumor Heterogeneity	Significant expression variability between biopsy/resection/lymph nodes	Possible misclassification of PD-L1 status [57]
Specimen Adequacy	Small biopsies may have <100 viable tumor cells	Compromised TPS accuracy [41]
Tissue Processing	Varying staining intensity across processors	Inter-laboratory variability [70]
Scoring Method	Digital vs. visual assessment differences	Altered patient classification [41]

The evidence suggests that PD-L1 expression is not uniform across different tumor sites or sampling timepoints, reflecting dynamic changes in the tumor microenvironment [57]. This supports the practice of testing the most recent specimen available, as it best represents the current tumor biology that will encounter the therapeutic agent.

Experimental Protocols and Methodologies

Multi-Specimen Comparison Study Design

A rigorous experimental design was employed to directly compare PD-L1 expression across different specimen types from the same patients, eliminating inter-patient variability [57].

Materials and Methods:

• Patient Cohort: 68 HNSCC cases (39 oropharynx, 29 oral cavity) with complete sets of preoperative biopsy, surgical resection, and metastatic lymph node samples [57]
• IHC Staining: PD-L1 IHC 22C3 pharmDx assay on 204 tissue sections (68 cases × 3 specimens) automated on Ventana BenchMark ULTRA platform [57]
• Digital Analysis: Whole-slide imaging with Philips Intellisite Pathology Solution followed by bioimage analysis using QuPath open-source platform [57]
• Cell Population Annotation: Manual classification of four distinct cell populations - tumor cells, immune cells, PD-L1-expressing tumor cells, and PD-L1-expressing immune cells [57]
• Statistical Analysis: Kruskal-Wallis test for comparing CPS and TPS across specimen types; Mann-Whitney U test for association with HPV status [57]

This comprehensive approach enabled direct comparison of PD-L1 expression in matched specimens, providing unique insights into spatial heterogeneity while controlling for assay variability through standardized staining and analysis protocols.

Figure 1: Experimental workflow for multi-specimen PD-L1 comparison study

Quantitative Continuous Scoring Methodology

Advanced computational pathology approaches have been developed to address limitations of visual PD-L1 scoring, particularly for heterogeneous specimen types.

PD-L1 Quantitative Continuous Scoring (QCS) Protocol:

• Sample Preparation: 768 whole slide images from the MYSTIC trial (NCT02453282) were analyzed, including samples from anti-PD-L1 (Durvalumab), combination therapy, and chemotherapy arms [41]
• Digital Analysis: Computer vision system for granular cell-level quantification of PD-L1 staining intensity in digitized whole slide images [41]
• Biomarker Derivation: Percentage of tumor cells with medium to strong staining intensity (PD-L1 QCS-PMSTC) with optimal cut-point determination [41]
• Statistical Optimization: Standardized two-sample linear rank statistics incorporating survival data to identify optimal cut-points for parameterized features [41]
• Validation: Comparison against visual scoring of %TC ≥50 using hazard ratios and prevalence calculations [41]

This methodology demonstrated that digital scoring could identify patient populations with comparable survival benefits to visual scoring but with increased prevalence (54.3% vs. 29.7%), potentially allowing more patients to benefit from immunotherapy [41].

Integrated Analysis and Complementary Biomarkers

Combined Biomarker Strategies

Given the limitations of PD-L1 assessment alone, particularly with variable specimen types, research has explored combination biomarker strategies to improve predictive accuracy.

Table 3: Combined PD-L1 and TILs Biomarker Performance

Biomarker Combination	PFS Improvement (HR)	OS Improvement (HR)	Number of Studies
PD-L1 alone	0.67 (CI: 0.49-0.90)	Not significant	8 evaluable studies [15]
TILs alone	Not significant	Not significant	8 evaluable studies [15]
PD-L1 + TILs combined	0.39 (CI: 0.27-0.57)	0.42 (CI: 0.31-0.56)	7 of 7 studies showed benefit [15]

The synergistic effect of combining PD-L1 with tumor-infiltrating lymphocytes (TILs) suggests that comprehensive assessment of the tumor immune microenvironment may compensate for limitations of individual biomarker assessment in specific specimen types [15].

Tissue Processing and Analytical Variability

The impact of pre-analytical variables on PD-L1 testing consistency across different specimen types cannot be overstated. A systematic evaluation of tissue processing demonstrated significant technical variability:

Tissue Processing Evaluation Protocol:

• Sample Selection: 73 samples across three topographies (uterine leiomyomas, placentas, palatine tonsils) [70]
• Experimental Design: Three fragments from each sample processed in different tissue processors (A, B, C) with color-coding for tracking [70]
• Staining Protocol: PD-L1 IHC with 22C3 and SP142 clones on sequential sections from same blocks [70]
• Quality Assessment: Blind evaluation by two pathologists using digital pathology platform [70]

Key Findings: Tissue processor C demonstrated 50.7% artifact incidence, and SP142 PD-L1 staining was considered inadequate for evaluation in 29.2% of cases after processing with this system, highlighting how technical processing variables interact with different antibody clones to affect result reliability [70].

Figure 2: Factors influencing PD-L1 expression variability and clinical impact

The Scientist's Toolkit

Table 4: Essential Research Reagent Solutions for PD-L1 Specimen Studies

Reagent/Platform	Function	Application Notes
PD-L1 IHC 22C3 PharmDx	Companion diagnostic antibody	FDA-approved for pembrolizumab; high concordance with SP263 [56] [71]
PD-L1 IHC SP263 Assay	Complementary diagnostic antibody	Ventana platform; comparable to CAL10 in development [56]
Ventana BenchMark ULTRA	Automated staining platform	Standardized IHC processing; reduces technical variability [57]
QuPath Bioimage Analysis	Digital pathology platform	Open-source solution for CPS/TPS calculation [57]
Aperio GT 450 Scanner	Whole slide imaging	High-resolution digitization for quantitative analysis [56]
FFPE Tissue Blocks	Specimen preservation	Standard material for PD-L1 IHC; enables archival studies [57]
NIST Traceable Calibrators	Assay standardization	Quantitative comparison across laboratories and assays [72]
Salmeterol-d3	Salmeterol-d3, CAS:497063-94-2, MF:C25H37NO4, MW:418.6 g/mol	Chemical Reagent
Atraric Acid	Methyl 2,4-dihydroxy-3,6-dimethylbenzoate|CAS 4707-47-5	Methyl 2,4-dihydroxy-3,6-dimethylbenzoate (Atraric Acid). High-purity grade for antiandrogen, fragrance, and organic synthesis research. For Research Use Only. Not for human or veterinary diagnostic or therapeutic use.

The comparative analysis of PD-L1 testing across different specimen types reveals substantial technical and biological challenges that impact biomarker reliability. Surgical resection specimens generally provide the most comprehensive assessment of PD-L1 expression but are not always available in advanced disease settings. Biopsies and metastatic lymph node specimens demonstrate significant variability in PD-L1 expression compared to matched resections, potentially leading to different treatment classifications. The integration of digital pathology solutions and standardized processing protocols can mitigate some variability, while combination biomarker approaches incorporating TILs may provide more robust predictive value across diverse specimen types. Researchers and clinicians should prioritize specimen quality standardization and consider implementing computational pathology solutions to improve consistency in PD-L1 assessment across different tissue specimens.

Navigating PD-L1 Testing Challenges: Pre-analytical Variables and Interpretation Pitfalls

In the era of precision oncology, the accurate assessment of biomarkers like Programmed Death-Ligand 1 (PD-L1) through immunohistochemistry (IHC) is crucial for identifying patients eligible for immune checkpoint inhibitor therapy. Pre-analytical factors—encompassing all procedures from tissue collection to antigen retrieval—represent critical variables that significantly influence the reliability and reproducibility of IHC results. Variations in fixation timing, tissue processing protocols, and storage conditions can profoundly affect epitope preservation, potentially leading to false-negative or false-positive interpretations that directly impact therapeutic decisions. For predictive biomarkers such as PD-L1, which guide treatment with agents like pembrolizumab in non-small cell lung cancer (NSCLC) and triple-negative breast cancer (TNBC), standardized pre-analytical workflows are not merely recommendations but essential components of quality assurance in pathology practice [73] [74].

The complexity of PD-L1 as a biomarker, with its dynamic expression patterns and multiple FDA-approved companion diagnostic assays, further underscores the necessity of controlling pre-analytical variables. Evidence indicates that suboptimal tissue handling can alter staining intensity and distribution, compromising the clinical utility of this critical biomarker. This guide systematically examines the experimental evidence quantifying the effects of key pre-analytical factors on PD-L1 IHC performance, providing researchers and drug development professionals with evidence-based protocols to ensure analytical validity in both research and diagnostic contexts.

Effects of Fixation Delay on Protein Epitope Integrity

Experimental Evidence from Controlled Studies

The interval between tissue resection and formalin fixation, termed cold ischemia time, represents one of the most critical pre-analytical variables affecting IHC quality. A systematic investigation into fixation delays utilized lung resection specimens from NSCLC tumors larger than 4 cm, collecting ten samples per case subjected to different fixation protocols [75]. Researchers created tissue microarrays (TMAs) from these samples and stained them with 20 different antibodies, including PD-L1 (clones 22C3 and E1L3N), scoring for staining quality and intensity using a standardized scoring system.

The experimental design included samples with delayed fixation (1 hour, 6 hours, 24 hours, 48 hours, and 96 hours) alongside standard fixation controls (0 hours delay) and prolonged fixation samples (2 days, 4 days, 7 days). This comprehensive approach allowed for direct comparison of fixation timing effects across multiple biomarkers relevant to lung cancer diagnosis and treatment [75].

Table 1: Effects of Fixation Delay on IHC Quality in NSCLC Tissue

Parameter Assessed	Findings with Delayed Fixation	Statistical Significance
TMA Core Loss	35% core loss vs. 27% in prolonged fixation	p<0.01 for multiple markers
Tissue Quality Deterioration	Significant reduction in interpretable cores	Score 5 (poor quality) increased
PD-L1 Expression	Reduction in immunoreactivity	Significant decrease
Cytokeratin Markers	Reduced expression of CK7, CAM 5.2, Keratin MNF116	Significant decrease
Diagnostic Markers	Reduced TTF-1, Napsin A, CK5/6 expression	Significant decrease

The findings demonstrated that delayed fixation negatively affected tissue morphology and antigen preservation, with samples experiencing fixation delays showing significant loss of TMA cores on glass slides and deterioration of tissue quality [75]. This resulted in measurable reduction in expression levels across multiple immunohistochemical markers, including those with diagnostic relevance (cytokeratins, TTF-1) and predictive value (PD-L1). In contrast, prolonged fixation (up to 7 days) showed no significant adverse effects on IHC performance, suggesting that extended formalin exposure is less detrimental than delayed fixation [75].

Mechanisms of Ischemia-Induced Epitope Degradation

The degradation of protein epitopes during delayed fixation occurs through multiple mechanisms. Tissue ischemia triggers enzymatic degradation pathways, including protease and phosphatase activation, which modify protein structure and compromise antibody binding sites. Additionally, oxidative damage and pH changes in non-fixed tissues can alter protein conformation, particularly affecting phosphorylation-dependent epitopes. The variation in sensitivity to ischemia across different markers reflects differences in epitope stability and the specific structural requirements for antibody recognition [73].

Tissue Processing and Storage Duration Effects

Impact of FFPE Block Storage on PD-L1 Immunoreactivity

The duration of formalin-fixed paraffin-embedded (FFPE) tissue block storage represents an underappreciated pre-analytical variable with significant implications for PD-L1 IHC validity. A recent investigation examined 63 triple-negative breast cancer cases with PD-L1 testing using the 22C3 pharmDx assay, evaluating immunoreactivity decline relative to storage duration [74]. The study employed a retrospective design, repeating PD-L1 IHC on the same FFPE blocks after varying storage intervals and comparing results with baseline assessments conducted at initial diagnosis.

Table 2: PD-L1 Immunoreactivity Decline with FFPE Block Storage

Storage Duration	Percentage Showing Decreased Staining	False-Negative Risk
<1 year	0%	Minimal
1-2 years	11%	Low
2-3 years	13%	Moderate
≥3 years	50%	High

The results demonstrated a striking time-dependent decline in PD-L1 immunoreactivity, with 50% of initially PD-L1-positive cases showing significantly reduced staining after three or more years of storage at room temperature [74]. This decline has direct clinical implications, as false-negative PD-L1 results could inappropriately exclude patients from potentially beneficial immune checkpoint inhibitor therapy. The study also identified associations between PD-L1 positivity and higher Ki67 proliferation index and nuclear grade, suggesting that particularly aggressive tumors might be disproportionately affected by storage-related false negatives [74].

Molecular Mechanisms of Storage-Related Epitope Degradation

The degradation of protein epitopes during FFPE block storage results from multiple molecular processes. Protein oxidation, hydrolysis, and continued cross-linking reactions gradually modify antigenic structures, compromising antibody binding affinity. Environmental factors such as storage temperature, humidity, and paraffin quality further influence degradation rates. The particular vulnerability of PD-L1 epitopes to storage conditions underscores the need for standardized archival protocols, especially for retrospective studies utilizing historical tissue blocks [74].

Comparative Performance of PD-L1 Detection Methodologies

Multiplexed Approaches for Enhanced Predictive Value

While standard PD-L1 IHC remains widely used for predicting response to immune checkpoint inhibitors, emerging evidence suggests that alternative methodologies may offer superior predictive performance. A comprehensive network meta-analysis compared seven different testing methodologies for predicting response to PD-1/PD-L1 inhibitors, analyzing 144 diagnostic index tests from 49 studies encompassing 5,322 patients [76].

Table 3: Comparative Performance of PD-L1 Detection Methodologies

Methodology	Sensitivity	Specificity	Diagnostic Odds Ratio	Best Application Context
Multiplex IHC/IF (mIHC/IF)	0.76	0.57-0.89	5.09	Pan-cancer
MSI	0.90	0.85-0.94	6.79	Gastrointestinal tumors
PD-L1 IHC + TMB	0.89	0.82-0.94	Not reported	NSCLC
Standard PD-L1 IHC	Variable by clone	Variable by clone	Lower than alternatives	Companion diagnostic context

The analysis revealed that multiplex IHC/immunofluorescence (mIHC/IF) exhibited the highest sensitivity (0.76) and second-highest diagnostic odds ratio (5.09), suggesting superior overall performance in predicting response to anti-PD-1/PD-L1 therapy [76]. Microsatellite instability (MSI) status demonstrated the highest specificity (0.90) and diagnostic odds ratio (6.79), particularly in gastrointestinal tumors. Notably, the combination of PD-L1 IHC with tumor mutational burden (TMB) significantly improved sensitivity to 0.89, indicating that integrated biomarker approaches may outperform single-analyte tests [76].

Artificial Intelligence-Enhanced PD-L1 Assessment

Recent advancements in computational pathology have introduced artificial intelligence (AI) approaches for PD-L1 assessment directly from hematoxylin and eosin-stained histological slides. Deep learning algorithms can predict PD-L1 expression patterns while reducing the interobserver variability associated with manual scoring methods like Tumor Proportion Score and Combined Positive Score [77]. These AI-driven tools offer potential for standardized, reproducible PD-L1 assessment while potentially mitigating some pre-analytical challenges through pattern recognition capabilities that may be less affected by subtle epitope degradation.

Standardized Experimental Protocols for Pre-analytical Variable Evaluation

Protocol for Evaluating Fixation Delay Effects

Experimental Design:

• Collect multiple matched tissue samples from surgical specimens (minimum 0.5 × 0.5 × 0.3 cm)
• Assign samples to different fixation delay intervals (0, 1, 6, 24, 48, 96 hours) at room temperature
• Include control samples with standard fixation (24 hours in 10% neutral buffered formalin)
• Process all samples through identical dehydration, clearing, and paraffin embedding protocols
• Construct tissue microarrays with multiple cores per sample to ensure statistical power

Assessment Methodology:

• Perform IHC with validated antibodies including PD-L1 (multiple clones), cytokeratins, and cell-specific markers
• Utilize standardized scoring systems incorporating intensity (0-4+) and quality (1-5) metrics
• Employ multiple blinded pathologists for scoring to minimize observer bias
• Include normal tissue elements as internal controls when possible
• Statistical analysis using McNemar and Wilcoxon signed rank tests for comparison [75]

Protocol for Assessing Storage Duration Effects

Experimental Design:

• Identify cases with previous PD-L1 testing at initial diagnosis
• Select FFPE blocks stored under documented conditions (temperature, humidity)
• Include blocks across storage duration categories (<1, 1-2, 2-3, ≥3 years)
• Section new slides from original blocks using identical microtomy protocols
• Process slides simultaneously using standardized IHC protocols identical to initial testing

Assessment Methodology:

• Evaluate PD-L1 expression using identical scoring systems (TPS, CPS) as initial assessment
• Directly compare current and historical results for paired samples
• Correlate staining changes with clinicopathological parameters
• Control for inter-batch variation using reference standards in each staining run
• Statistical analysis using appropriate tests (e.g., Fisher's exact, Cochran-Armitage trend test) [74]

The Scientist's Toolkit: Essential Research Reagent Solutions

Table 4: Essential Research Reagents for Pre-analytical Studies

Reagent/Category	Specific Examples	Research Function
Fixatives	10% Neutral Buffered Formalin, Zinc Formalin	Tissue preservation and antigen stabilization
PD-L1 Antibody Clones	22C3, 28-8, SP142, SP263	Detection of PD-L1 expression with clone-specific characteristics
IHC Detection Systems	Polymer-based detection, chromogenic substrates	Signal amplification and visualization
Tissue Processing Reagents	Ethanol, xylene, low-melt paraffin	Tissue dehydration, clearing, and embedding
Antigen Retrieval Solutions	Citrate buffer (pH 6.0), EDTA/TRIS (pH 9.0)	Epitope unmasking through heat-induced methods
Control Materials	Cell line arrays, multitissue blocks	Assay validation and quality control

Signaling Pathways and Experimental Workflows

PD-1/PD-L1 Signaling Pathway and IHC Detection Implications

Diagram 1: The PD-1/PD-L1 signaling pathway and its relationship to pre-analytical factors in IHC detection. This pathway illustrates how T-cell-derived IFN-γ induces PD-L1 expression on tumor cells, leading to T-cell inhibition upon binding. Immune checkpoint inhibitors block this interaction, restoring anti-tumor immunity. Pre-analytical variables directly impact PD-L1 detection accuracy, which informs treatment decisions.

Experimental Workflow for Pre-analytical Variable Assessment

Diagram 2: Comprehensive experimental workflow for assessing pre-analytical variables in PD-L1 IHC. This workflow systematically evaluates the impact of fixation delay, storage duration, and antigen retrieval methods while controlling for consistent fixative type, storage conditions, and antibody clones. The standardized approach enables quantitative comparison of staining quality across experimental conditions.

The cumulative evidence from controlled studies demonstrates that pre-analytical factors—particularly fixation delay and storage duration—significantly impact PD-L1 IHC reliability, with potential consequences for patient selection in immunotherapy. Fixation delays exceeding 6-12 hours and FFPE block storage beyond three years substantially reduce immunoreactivity, potentially leading to false-negative results that exclude patients from beneficial treatments. Conversely, prolonged fixation appears less detrimental, while emerging methodologies like multiplex IHC/IF and AI-assisted analysis show promise for enhanced predictive performance.

For researchers and drug development professionals, these findings underscore the necessity of implementing standardized pre-analytical protocols across institutions. Specific recommendations include minimizing cold ischemia time to under 1 hour when possible, establishing storage duration limits for FFPE blocks used in biomarker studies, and adopting multiplexed approaches where feasible. Future efforts should focus on developing stabilization technologies resistant to pre-analytical variability and establishing universal quality metrics for tissue processing in predictive biomarker analysis. Through rigorous attention to pre-analytical variables, the field can improve the reproducibility and clinical utility of PD-L1 immunohistochemistry in precision oncology.

The accurate assessment of programmed death-ligand 1 (PD-L1) expression via immunohistochemistry (IHC) is a critical component of precision oncology, serving as a primary biomarker for predicting responses to immune checkpoint inhibitors (ICIs) [20]. However, the diagnostic landscape is complicated by the existence of multiple, commercially available PD-L1 IHC assays and scoring algorithms, leading to significant challenges in standardization and interpretation. Two major sources of variability—inter-assay differences and inter-observer concordance—directly impact the reliability of PD-L1 testing and, consequently, patient selection for immunotherapy.

Inter-assay variability arises from the use of different antibody clones, staining platforms, and scoring criteria, which can yield discordant results for the same tumor sample [78]. Simultaneously, inter-observer variability reflects the challenges pathologists face in consistently interpreting IHC stains, particularly for complex scoring systems that evaluate both tumor and immune cells [79]. This guide objectively compares the performance of major PD-L1 assays, summarizes key experimental data on concordance, and details the methodologies used to generate this evidence, providing researchers and clinicians with a clear framework for evaluating assay performance in a regulatory and research context.

Key Comparative Data on PD-L1 Assays

Data from multiple comparability studies reveal consistent patterns of performance and concordance among the most widely used PD-L1 IHC assays. The findings are synthesized in the tables below.

Table 1: Inter-Assay Analytical Concordance Across Multiple Tumor Types

Assay Comparison	Tumor Type	Scoring Method	Concordance Level	Key Findings
22C3 vs 28-8 vs SP263	Triple-Negative Breast Cancer (TNBC)	IC-score ≥1% / CPS ≥1	Good Agreement (κ 0.68-0.74 for 22C3/28-8) [80]	22C3, 28-8, and SP263 showed comparable positivity rates; SP263 was not interchangeable with others for all scores [80].
22C3 vs 28-8 vs SP263	Hepatocellular Carcinoma (HCC)	TPS / CPS	Highly Concordant [17]	These three assays demonstrated high concordance, suggesting potential interchangeability [17].
SP142 vs others	TNBC and HCC	IC-score / CPS	Lower Sensitivity & Concordance [80] [17]	SP142 was consistently the least sensitive assay, with lower positivity rates and concordance with other assays [80] [17].
CAL10 vs SP263	Non-Small Cell Lung Cancer (NSCLC)	TPS ≥1% / ≥50%	High Concordance (OPA ≥94% at TPS≥1%) [56]	The novel CAL10 assay demonstrated comparable performance to the SP263 assay, meeting pre-defined concordance targets [56].

Table 2: Inter-Observer Agreement for PD-L1 Scoring

Assay	Tumor Type	Scoring Method & Cut-off	Inter-Observer Agreement	Intra-Observer Agreement
Four Assays (SP142, SP263, 22C3, 28-8)	TNBC	IC-score ≥1% / CPS ≥1	Good to Excellent (κ 0.73-0.78) [80]	Not Reported
SP142	Breast Cancer (Various Subtypes)	IC-score ≥1%	Substantial (Fleiss κ 0.654-0.655) [79]	Substantial to Almost Perfect (κ 0.667-0.956) [79]
SP142	Breast Cancer (Various Subtypes)	IC-score (Continuous)	ICC: Good to Excellent (Overall) [79]	ICC: Good to Excellent [79]
Four Assays (SP142, SP263, 22C3, 28-8)	HCC	TPS / CPS	Good to Excellent (ICC TPS: 0.946; CPS: 0.809) [17]	Not Reported

Experimental Protocols for Key Studies

The comparative data presented above are derived from rigorously designed studies. The methodologies of the most comprehensive investigations are detailed below.

Protocol: Large-Scale Multi-Assay Comparison in TNBC

A 2021 study directly compared four clinically relevant PD-L1 assays in a cohort of 104 triple-negative breast cancer resection specimens [80].

• Sample Preparation: Archival, formalin-fixed, paraffin-embedded (FFPE) resection specimens were used. The VENTANA SP142 and SP263 assays were stained on the VENTANA Benchmark Ultra platform, while the DAKO 22C3 and 28-8 assays were run on a DAKO Autostainer Link 48 [80].
• Digital Slide Analysis: All stained slides and corresponding H&E stains were digitized using a Leica Aperio AT2 slide scanner to create virtual whole slide images for analysis [80].
• Blinded Scoring: Four trained pathologists evaluated the digital slides in a randomized and blinded manner. Each pathologist scored PD-L1 expression for immune cells (IC) and tumor cells (TC) separately, calculating the Immune Cell score (IC-score), Tumor Proportion Score (TPS), and Combined Positive Score (CPS) according to their respective guidelines [80].
• Statistical Analysis: Inter-assay agreement was quantified using Fleiss' Kappa for categorical agreement (positive/negative at specific cut-offs). Inter-observer agreement was also assessed using Fleiss' Kappa and Intraclass Correlation Coefficients (ICC) for continuous measurements [80].

Protocol: Inter-Observer Variability of SP142 in Breast Cancer

A 2023 study specifically evaluated the inter- and intra-observer agreement of the SP142 assay in a multi-subtype breast cancer cohort [79].

• Cohort and Staining: The study included 100 core biopsies from primary breast cancers, enriched for TNBC but including luminal and HER2-positive subtypes. All samples were stained with the VENTANA SP142 assay on a VENTANA Benchmark ULTRA platform [79].
• Digital Scoring Rounds: The stained slides were digitally scanned and uploaded to an online platform. Twelve pathologists from multiple institutions independently scored the IC-score for all cases. A subset of ten pathologists re-scored the same cases after a washout period of at least three months to assess intra-observer variability [79].
• Statistical Metrics: Absolute agreement and consensus scoring were calculated. Inter- and intra-observer agreements for the binary outcome (positive/negative at IC-score ≥1%) were measured using Fleiss' Kappa. The reliability of continuous percentage scores was assessed using the Intraclass Correlation Coefficient (ICC) [79].

Visualizing Assay Concordance and Variability

The relationships between different assays and the sources of variability in PD-L1 testing can be visualized through the following diagrams.

Assay Variability and Concordance Factors

This diagram illustrates the primary factors contributing to inter-assay and inter-observer variability in PD-L1 testing, and how they influence the final, harmonized result.

Typical PD-L1 Comparability Study Workflow

This diagram outlines the standard workflow for a PD-L1 assay comparability study, from sample processing to statistical analysis of concordance.

The Scientist's Toolkit: Key Reagents and Platforms

Table 3: Essential Research Reagents and Platforms for PD-L1 IHC

Item	Function in PD-L1 Research	Examples / Notes
Antibody Clones	Primary antibodies that specifically bind to the PD-L1 epitope. Different clones have varying sensitivities and specificities.	22C3, 28-8 (Agilent); SP142, SP263 (Roche); CAL10 (Leica) [80] [78] [56].
Automated Staining Platforms	Automated IHC instruments that standardize the staining process to reduce technical variability.	DAKO Autostainer Link 48 (for 22C3, 28-8); VENTANA Benchmark Ultra (for SP142, SP263); BOND-III (for CAL10) [80] [56].
Digital Pathology System	High-resolution slide scanners and software for creating, storing, and analyzing whole slide images (WSIs).	Enables blinded, remote review by multiple pathologists and facilitates computational analysis [80] [79] [56].
Positive Control Tissues	Tissues with known PD-L1 expression levels used to validate staining run performance.	Multi-tissue blocks containing tonsil and placenta are commonly used [56]. Cell line blocks can also serve as controls [79].
Statistical Analysis Tools	Software for calculating concordance metrics and reliability statistics.	R or SPSS software for calculating Fleiss' Kappa, Intraclass Correlation Coefficient (ICC), and Overall Percent Agreement (OPA) [80] [79] [56].

The accurate assessment of programmed death-ligand 1 (PD-L1) expression via immunohistochemistry (IHC) represents a critical cornerstone in predicting response to immune checkpoint blockade (ICB) therapy. However, this process is fundamentally complicated by extensive spatial and temporal tumor heterogeneity, which introduces significant variability in biomarker interpretation. These heterogeneity issues manifest as varying PD-L1 expression patterns across different tumor regions, between primary and metastatic sites, and throughout disease progression and treatment. Consequently, limited biopsy samples may fail to capture the complete immunological landscape of a tumor, potentially leading to misclassification of a patient's PD-L1 status and suboptimal treatment decisions. This comparative analysis examines how spatial heterogeneity, assay concordance, and novel detection methodologies influence the reliability of PD-L1 testing, providing researchers and drug development professionals with a structured evaluation of current challenges and emerging solutions in the field of comparative immunohistochemistry assay performance.

Spatial Heterogeneity and Its Impact on PD-L1 Assessment

Spatial heterogeneity of PD-L1 expression presents a substantial obstacle for reliable biomarker evaluation in esophageal squamous cell carcinoma (ESCC) and other solid tumors. Research demonstrates that PD-L1 expression exhibits significant intratumoral spatial heterogeneity, which can render limited biopsy samples unrepresentative of the overall tumor PD-L1 status [81].

Experimental Evidence of Spatial Heterogeneity

A prospective observational study employed rigorous methodology to quantify spatial heterogeneity in treatment-naïve ESCC patients. The experimental protocol involved:

• Multi-region Sampling: For cohort 1 (n=30), four distinct tumor regions larger than 3mm each were sampled using endoscopic biopsy forceps from surgically resected tumors: proximal tumor region (A), distal tumor region (B), surface mid-region (C), and tumor center (D), with regions selected at least 0.5cm apart while avoiding areas of severe necrosis [81].

• Complete Tumor Specimen Analysis: For cohort 2 (n=4), the largest longitudinal section along the midline of completely resected tumors was divided into 3mm × 3mm regions for comprehensive analysis, with regions included only if they comprised at least one-third tissue area and tumor cells accounted for at least one-third of the tissue area [81].

• PD-L1 Quantification: PD-L1 expression was calculated using Combined Positive Score (CPS), defined as the number of PD-L1 stained cells (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells multiplied by 100, with a minimum requirement of 100 viable tumor cells per evaluation area [81].

Table 1: Spatial Heterogeneity of PD-L1 Expression in ESCC

Assessment Method	Regional Discordance Rate	Key Findings	Reduction Strategy
Multi-region biopsy (4 regions)	Significant regional discordance observed	Limited biopsies often unrepresentative of bulk tumor	Multi-region sampling (3 regions)
Complete tumor specimens	Variation across normalized regions	Heterogeneity reduced with sufficiently high CPS	Maximum CPS from multiple regions
Correlation with T-cells	N/A	CPS positively correlated with CD8+/CD4+ T-cell density	Standardized biopsy strategy

Relationship Between PD-L1 Expression and Immune Context

The spatial distribution of PD-L1 expression demonstrates a significant correlation with the tumor immune microenvironment. Quantitative analyses of CD8+ and CD4+ T-cell infiltration densities, performed using immunohistochemistry on serial sections from the same field of view used for CPS assessment, revealed positive correlations between CPS and T-cell densities [81]. This relationship underscores the biological connection between immune cell infiltration and PD-L1 upregulation, while simultaneously highlighting how spatial heterogeneity in immune cell distribution can consequently drive heterogeneous PD-L1 expression patterns.

Comparative Performance of FDA-Approved PD-L1 Assays

The concordance between different FDA-approved PD-L1 immunohistochemistry assays varies significantly across cancer types, presenting challenges for standardized biomarker implementation. A comprehensive evaluation of four PD-L1 assays in clear cell renal cell carcinoma (ccRCC) revealed substantial differences in detection capabilities and prognostic value [25].

Experimental Protocol for Assay Comparison

The methodological approach for comparative assay assessment included:

• Tissue Microarray Construction: Researchers constructed TMAs from 286 ccRCC tissue samples, enabling standardized evaluation across multiple specimens under uniform conditions [25].

• Parallel IHC Staining: Each sample was evaluated using four FDA-approved PD-L1 assays: 22C3, 28-8, SP142, and SP263, with strict adherence to the specific evaluation criteria established for each assay in their respective clinical trials [25].

• Assessment Criteria: Evaluation included PD-L1 expression in tumor cells (TC), immune cells (IC), and combined scores where applicable, following manufacturer specifications and clinical trial protocols [25].

• Concordance Analysis: Pairwise concordance was assessed using kappa statistics, with prognostic correlation evaluated through cancer-specific survival analysis [25].

Table 2: Performance Comparison of FDA-Approved PD-L1 Assays in ccRCC

Assay	PD-L1+ in Tumor Cells	PD-L1+ in Immune Cells	Concordance (κ with 28-8)	Prognostic Value
22C3	18.9%	14.7%	0.52 (TC)	Worse CSS with IC+
28-8	2.1%	16.1%	Reference	Worse CSS with IC+
SP142	2.1%	2.1%	0.16 (IC)	Limited prognostic value
SP263	15.0%	15.0%	0.46 (IC)	Worse CSS with combined+

Key Findings on Assay Variability

The comparative analysis revealed that PD-L1 expression in tumor cells was generally low across all assays in ccRCC, while expression in immune cells showed greater variability, approximately 15% for most assays except SP142, which demonstrated remarkably low positivity [25]. The 28-8 assay showed the highest agreement with other assays, while SP142 was deemed unsuitable for concordance evaluation due to its exceptionally low detection rate [25]. Critically, patients with PD-L1 expression in immune cells assessed using 22C3, 28-8, and SP263 assays showed significantly worse cancer-specific survival, highlighting the clinical implications of assay selection [25].

The following diagram illustrates the experimental workflow for comparative PD-L1 assay evaluation:

Diagram 1: PD-L1 Assay Comparison Workflow. This workflow illustrates the experimental protocol for evaluating four FDA-approved PD-L1 assays using tissue microarrays from clear cell renal cell carcinoma patients.

Emerging Solutions and Novel Approaches

Heterogeneity-Optimized Computational Frameworks

Advanced computational approaches are being developed to address the challenges posed by tumor heterogeneity in ICB response prediction. A novel heterogeneity-optimized machine learning framework demonstrates how addressing multimodal distribution in cancer data can enhance prediction accuracy [82].

The methodological framework involves:

• Heterogeneity-Aware Clustering: Application of K-means clustering (K=2) to stratify patients into biologically distinct "hot-tumor" and "cold-tumor" subgroups based on multimodal tumor data, outperforming hierarchical clustering and DBSCAN alternatives [82].

• Subtype-Specific Modeling: Development of separate predictive models for each subgroup—a support vector machine for hot-tumor subtypes and a random forest for cold-tumor subtypes—utilizing seven heterogeneity-associated biomarkers to overcome unimodal distribution assumptions [82].

• Validation: The framework demonstrated enhanced ICB response prediction across melanoma, NSCLC, and pan-cancer datasets, achieving a mean accuracy gain of at least 1.24% compared to 11 baseline methods, with consistent performance in independent external validation cohorts [82].

Exosomal PD-L1 as a Liquid Biopsy Alternative

Exosomal PD-L1 (exo-PD-L1) has emerged as a promising solution to spatial sampling limitations, offering a systemic, minimally invasive biomarker that captures immune status across tumor sites [83].

The biogenesis and function of exosomal PD-L1 involves:

• Formation: Exo-PD-L1 originates from plasma membrane endocytosis, forming early endosomes that develop into multivesicular bodies (MVBs) containing intraluminal vesicles, which are released as exosomes (30-150nm) upon MVB fusion with the cellular membrane [83].

• Function: Exo-PD-L1 retains membrane topology, enabling PD-1 binding on T cells and systemic immunosuppression through inhibition of PI3K-AKT and MAPK pathways, restriction of T cell proliferation, and promotion of T cell exhaustion and senescence [83].

• Regulation: Interferon-gamma significantly stimulates exo-PD-L1 release as an adaptive immune evasion mechanism, with levels varying by cancer type (lower in "cold" tumors like ovarian cancer, higher in "hot" tumors like melanoma and NSCLC) [83].

The following diagram illustrates the mechanism of exosomal PD-L1 biogenesis and function:

Diagram 2: Exosomal PD-L1 Biogenesis and Function. This diagram illustrates the formation and immunosuppressive mechanism of exosomal PD-L1, from cytokine-stimulated biogenesis to systemic T-cell inhibition.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Key Research Reagent Solutions for PD-L1 Heterogeneity Studies

Reagent/Assay	Manufacturer	Primary Function	Application Context
VENTANA PD-L1 (SP263) Assay	Roche Diagnostics	PD-L1 IHC detection	Companion diagnostic for multiple ICIs; demonstrated cost-effectiveness in NSCLC [84]
Dako PD-L1 IHC 22C3 pharmDx	Agilent Technologies	PD-L1 IHC detection	Companion diagnostic for pembrolizumab; used in comparative concordance studies [25]
Dako PD-L1 IHC 28-8 pharmDx	Agilent Technologies	PD-L1 IHC detection	Used in comparative assays; showed highest concordance with other tests [25]
VENTANA PD-L1 (SP142) Assay	Roche Diagnostics	PD-L1 IHC detection	Demonstrated low positivity in ccRCC; limited prognostic value [25]
Tissue Microarray Platform	Multiple vendors	High-throughput tissue analysis	Standardized evaluation of multiple specimens under uniform conditions [25]
Exosome Isolation Kits	Multiple vendors	Exo-PD-L1 enrichment	Liquid biopsy approach for systemic PD-L1 assessment [83]
Single-cell RNA Sequencing Kits	10X Genomics, etc.	Tumor microenvironment deconvolution	Identification of immune cell subtypes and spatial ecotypes [85]

The comparative analysis of PD-L1 immunohistochemistry assays reveals that tumor heterogeneity presents substantial challenges for reliable biomarker assessment, with spatial variation significantly impacting the accuracy of PD-L1 evaluation in esophageal squamous cell carcinoma and other solid tumors. The concordance between FDA-approved assays varies considerably across cancer types, with the SP263 and 22C3 assays generally demonstrating better performance characteristics compared to the SP142 assay in clear cell renal cell carcinoma. Emerging solutions including heterogeneity-optimized computational frameworks, exosomal PD-L1 detection, and standardized multi-region sampling protocols offer promising approaches to overcome these limitations. For researchers and drug development professionals, these findings emphasize the critical importance of selecting appropriate detection methodologies, implementing rigorous sampling strategies, and interpreting PD-L1 expression results within the context of tumor heterogeneity to optimize immunotherapy prediction and patient stratification.

Optimization of Staining Protocols Across Different Automated Platforms

Programmed death-ligand 1 (PD-L1) immunohistochemistry (IHC) is a critical predictive biomarker for identifying patients eligible for immune checkpoint inhibitor therapy across multiple cancer types, including non-small cell lung cancer (NSCLC) and gastric cancer [58] [71]. The accurate assessment of PD-L1 expression via the Combined Positive Score (CPS) or Tumor Proportion Score (TPS) is essential for optimal patient selection. However, the existence of multiple automated staining platforms, different antibody clones, and varied staining protocols introduces significant pre-analytical variability that can compromise the reliability and reproducibility of PD-L1 scoring [71]. This comparison guide objectively evaluates the performance of different automated platforms and staining protocols, providing researchers and drug development professionals with experimental data to optimize their PD-L1 detection systems.

Comparative Performance of PD-L1 IHC Assays

Key Automated Platforms and Assays

The landscape of automated PD-L1 IHC testing is dominated by two major staining platforms with their associated assays. The Dako Autostainer Link 48 platform runs the 22C3 and 28-8 assays, while the Ventana BenchMark series supports the SP263 and SP142 assays [71] [7]. Each assay was initially developed as a companion diagnostic for specific immune checkpoint inhibitors, creating practical challenges for laboratories that may need to run multiple tests.

Interassay Concordance and Variability

Multiple studies have systematically evaluated the concordance between different PD-L1 IHC assays to determine their potential interchangeability. The evidence reveals a complex pattern of agreement and divergence.

Table 1: Interassay Concordance in NSCLC PD-L1 Evaluation

Compared Assays	Concordance Level	Key Findings	Clinical Implications
22C3 vs 28-8 vs SP263	High agreement	Properly validated LDTs show strong correlation	Potential for interchangeability with proper validation
SP142 vs others	Lower concordance	Consistently shows different staining characteristics	Not recommended as substitute for other assays
All assays with cut-offs	Decreased concordance	Particularly problematic at 1% clinical decision threshold	Hampers interchangeability; requires platform-specific validation

A systematic review of 27 studies concluded that while high agreement exists between 22C3, 28-8, and SP263 assays, concordance decreases significantly when applying clinical cut-offs, particularly at the critical 1% threshold used for treatment decisions [71]. This finding highlights the challenges in assay interchangeability despite overall staining similarity.

The prospective study by Katayama et al. directly compared three different anti-PD-L1 antibodies (22C3, 28-8, and SP142) in 70 patients with advanced NSCLC treated with combined chemoimmunotherapy [7]. This research revealed that PD-L1 expression levels determined using the 22C3 assay showed the highest correlation with therapeutic response, successfully stratifying patients based on progression-free survival, while the other assays did not reveal remarkable differences in objective response rate or survival [7].

Experimental Data on Staining Protocol Performance

Impact on Clinical Outcomes

The optimization of staining protocols extends beyond technical performance to直接影响临床结果. In the Katayama study, only the 22C3 assay could significantly differentiate patient outcomes: those with TPS ≥50% showed significantly longer progression-free survival compared to those with TPS <50% [7]. This finding underscores how staining protocol selection can influence clinical decision-making quality.

Interobserver Concordance and Scoring Reproducibility

Staining protocol optimization must also consider the impact on interpretation consistency. The systematic review by PMC7318295 revealed that while interobserver concordance is generally high for all assays, agreement decreases significantly at the 1% cut-off [71]. This is particularly problematic in clinical practice, as discordance between pathologists at this threshold may result in eligible patients being denied valuable treatment options.

Table 2: Quantitative Performance Metrics of Automated PD-L1 Scoring Systems

Evaluation Metric	Manual Scoring Performance	AI-Assisted Scoring Performance	Improvement
Interobserver Agreement (ICC)	62%	74%	+12%
Agreement in Challenging Cases (CPS <20)	19%	62%	+43%
Classification Accuracy	75%	88%	+13%
Sensitivity	78%	96%	+18%
Positive Predictive Value	87%	88%	+1%

Recent advances in artificial intelligence (AI) have demonstrated potential to mitigate staining and interpretation variability. A clinical evaluation of the DiaKwant PD-L1 algorithm showed that AI assistance significantly improved interobserver agreement among pathologists, particularly in challenging cases with CPS <20 where ICC improved from 19% to 62% [86]. This demonstrates how computational approaches can complement staining protocol optimization.

Methodologies for Staining Protocol Evaluation

Standardized Experimental Workflow

To ensure fair comparison across different automated platforms, researchers have established standardized evaluation methodologies. The typical workflow involves parallel staining of identical tissue samples across different platforms, followed by blinded assessment by multiple pathologists, often supplemented with computational tools.

Automated Scoring Framework Methodology

For AI-assisted evaluation of staining protocol performance, advanced computational frameworks have been developed. These typically employ a multi-stage approach that combines machine learning models for tissue classification, segmentation, and cell detection.

The deep learning framework described by PMC12499557 utilizes a sophisticated pipeline where Vision Transformer-based models achieve 97.54% F1-score in tumor patch classification, while modified DeepLabV3+ architectures attain an 83.47% Dice Similarity Coefficient in tumor region segmentation [87]. This approach demonstrates remarkably high correlation (0.96) with pathologist-derived TPS scores, providing a robust methodological foundation for standardized staining protocol evaluation.

The Scientist's Toolkit: Essential Research Reagents and Materials

Optimizing staining protocols requires specific reagents and platforms designed for automated IHC staining. The following table details essential solutions for researchers conducting comparative studies of automated PD-L1 detection platforms.

Table 3: Essential Research Reagent Solutions for PD-L1 IHC Optimization

Reagent/Platform	Manufacturer	Primary Function	Application Notes
PD-L1 IHC 22C3 PharmDx	Dako/Agilent	Companion diagnostic for pembrolizumab	Optimized for Dako Autostainer Link 48 platform
PD-L1 IHC 28-8 PharmDx	Dako/Agilent	Complementary diagnostic for nivolumab	Shows high concordance with 22C3 and SP263
VENTANA PD-L1 (SP263)	Ventana/Roche	Companion diagnostic for durvalumab	CE marked for pembrolizumab and nivolumab identification
VENTANA PD-L1 (SP142)	Ventana/Roche	Complementary diagnostic for atezolizumab	Known for lower tumor cell staining intensity
BenchMark Special Stains	Ventana/Roche	Automated special stains platform	Enables customizable protocols with ready-to-use reagents
Dako Autostainer Link 48	Dako/Agilent	Automated IHC staining platform	Standardized staining for 22C3 and 28-8 assays
Hydrogen Peroxide (10%)	Various	Melanin bleaching agent	Critical for pigmented specimens; use at 60°C for 25 min
Alkaline Phosphatase (AP)	Various	Chromogenic detection	Superior contrast for melanin-rich specimens vs. DAB

The BenchMark Special Stains system exemplifies platform optimization, offering fully automated baking, deparaffinization, and staining with independent slide heating and customizable protocols to minimize technical variability [88]. For challenging specimens such as melanin-rich cytology samples, optimized bleaching protocols using 10% hydrogen peroxide at 60°C for 25 minutes significantly improve visualization without compromising cellular morphology or antigenicity [89].

Optimization of staining protocols across different automated platforms requires careful consideration of multiple performance factors. The experimental data demonstrates that while high concordance exists between 22C3, 28-8, and SP263 assays, significant differences emerge at clinical decision thresholds, particularly the 1% cut-off critical for treatment eligibility. The 22C3 assay shows particular utility for predicting response to chemoimmunotherapy in NSCLC patients. Integration of AI-assisted scoring systems significantly improves interobserver concordance, especially in challenging cases with low PD-L1 expression levels. Future optimization efforts should focus on standardizing pre-analytical variables, validating platform-specific cut-offs, and incorporating computational tools to maximize staining consistency and scoring reproducibility across different automated platforms.

Addressing Borderline Cases and Difficult-to-Interpret Staining Patterns

The accurate assessment of Programmed Death-Ligand 1 (PD-L1) expression through immunohistochemistry (IHC) is essential for identifying patients with cancer who may benefit from immune checkpoint inhibitor therapy [56] [71]. However, the existence of borderline cases—samples with PD-L1 expression levels near established clinical cut-offs—and challenging staining patterns presents significant interpretation difficulties for pathologists [56] [41]. These challenges contribute to substantial interobserver variability and may affect patient treatment eligibility [77] [71].

This guide objectively compares the performance of various PD-L1 IHC assays and emerging technologies in addressing these complex scenarios, providing researchers and drug development professionals with experimental data and methodologies to enhance assay selection and interpretation protocols.

Comparative Performance of PD-L1 IHC Assays

Analytical Concordance Across Platforms

Multiple studies have systematically evaluated the concordance between different PD-L1 IHC assays, particularly focusing on samples with expression levels near critical clinical thresholds. The following table summarizes key comparative performance data from recent studies.

Table 1: Comparative Performance of PD-L1 IHC Assays in NSCLC

Compared Assays	TPS Cut-off	Overall Percent Agreement (OPA)	Lower Bound of 95% CI	Sample Size (N)	Reference
CAL10 vs. SP263	≥50%	>86.2%*	86.2%	136	[56]
CAL10 vs. SP263	≥1%	>94.0%*	94.0%	136	[56]
22C3 vs. 28-8 vs. SP263	Various	High agreement	Not specified	Systematic Review	[71]
22C3 vs. 28-8 vs. SP263	≥1%	Lower concordance	Not specified	Systematic Review	[71]

Note: Exact OPA values not provided in source; lower bounds of 95% confidence intervals reported instead [56].

A systematic review of 27 studies confirmed high interassay concordance for the 22C3, 28-8, and SP263 assays, while properly validated laboratory-developed tests (LDTs) also demonstrated strong agreement with these standardized assays [71]. However, concordance decreases significantly when applying cut-offs, particularly at the 1% threshold, potentially impacting interchangeability in clinical practice [71].

Metrological Approaches to Assay Standardization

Fundamental differences in analytical sensitivity between PD-L1 assays contribute to discordance in borderline cases. A survey of 41 laboratories utilizing PD-L1 calibrators traceable to National Institute of Standards and Technology (NIST) Standard Reference Material 1934 revealed that the four FDA-cleared PD-L1 assays represent three distinct levels of analytical sensitivity [72].

Table 2: Metrological Characteristics of PD-L1 Assays

Assay Characteristic	Finding	Impact on Borderline Cases
Analytic Sensitivity	Varies between assays	Samples near cut-offs may be classified differently
Lower Limit of Detection (LOD)	Assay-dependent	Explains positive/negative discrepancies between assays
Dynamic Range	Disparate between some assays	Previous harmonization attempts unsuccessful for certain assays
LDT Performance	Some indistinguishable from predicate devices	Proper validation critical for reliable results

This metrological approach explains why previous attempts to harmonize certain PD-L1 assays proved unsuccessful—their dynamic ranges were too disparate and did not adequately overlap [72]. The implementation of standardized calibrators traceable to NIST standards represents an important transition for companion diagnostic testing that could improve patient stratification and test harmonization [72].

Experimental Protocols for Assay Comparison

Feasibility Study Protocol: CAL10 vs. SP263 Assays

A recently published feasibility study exemplifies a rigorous methodological approach for comparing novel PD-L1 assays with established platforms [56].

Sample Characteristics and Inclusion Criteria:

• Tissue Type: 136 formalin-fixed paraffin-embedded (FFPE) non-small cell lung cancer (NSCLC) tissue samples
• Histology Distribution: 60-70% adenocarcinomas, 30-40% squamous cell carcinomas, with inclusion of at least one large cell carcinoma sample
• Sample Origins: Both primary and metastatic sites represented
• Borderline Cases: Specifically included 21 cases with TPS in the 40-60% range (borderline for ≥50% cut-off)

Staining and Reading Procedures:

• Staining Platforms: CAL10 assay processed on BOND-III staining system (Leica Biosystems); SP263 assay processed on Benchmark Ultra system (Ventana)
• Control Tissues: Multi-tissue block containing tonsil and placenta tissues used as positive control
• Pathologist Review: Two pathologists independently read paired slide sets and recorded PD-L1 TPS on standardized score sheets
• Randomization: Stained NSCLC samples were randomized with unique IDs to minimize bias

Digital Pathology Integration:

• Scanning: CAL10-stained glass slides scanned using Aperio GT 450 scanner (Leica Biosystems)
• Digital Reading: After a 4-month washout period, whole slide images of CAL10-stained cases were read by pathologists
• Concordance Assessment: 130 cases used to assess comparability between manual and digital reading modalities

Statistical Analysis:

• A one-sided, exact, non-inferiority test for a single proportion with a 0.05 type 1 error rate was applied for both ≥50% and ≥1% TPS cut-offs
• Concordance was evaluated by assessing agreement rates between assays against PD-L1 scoring status (positive/negative) at each cut-off

Quantitative Continuous Scoring Protocol

Emerging methodologies address borderline cases through granular, continuous scoring systems that move beyond traditional categorical thresholds [41].

Sample Processing and Quality Control:

• Sample Source: 768 whole slide images from the MYSTIC trial (NCT02453282)
• Quality Assessment: 72 samples removed due to insufficient quality (inadequate tissue, staining artifacts, scanning issues)
• Technical Exclusions: 32 additional images excluded for scoring discrepancies (19 with out-of-focus regions, 13 with inappropriate image analysis)
• Final Cohort: Biomarker evaluable population of 768 samples

Quantitative Continuous Scoring Workflow:

Diagram 1: Quantitative continuous scoring workflow.

Biomarker Development and Validation:

• Classifier Optimization: Progression-free survival within the biomarker-positive group was maximized with constraint that log-rank p-value remained significant while maintaining 20-80% prevalence
• Positivity Threshold: Positive cells defined as having PD-L1 membrane staining intensity ≥40 (on a continuous scale)
• Sample Positivity: Sample considered positive when >3% of cells met staining intensity threshold
• Final Biomarker: Resulting score (PD-L1 QCS-PMSTC) represents proportion of medium to strongly stained tumor cells

Statistical Analysis for Cut-point Identification:

• Standardized two-sample linear rank statistics incorporating survival data (Log-Rank statistics) computed over reasonable cut-points
• Optimal cut-point determined as the one maximizing computed rank statistics for each treatment arm separately
• Spearman rank correlation used to select representative thresholds

Advanced Solutions for Chall staining Patterns

Artificial Intelligence and Computational Pathology

Recent advancements in artificial intelligence (AI) have transformed PD-L1 assessment approaches, particularly for borderline and difficult-to-interpret cases [77]. AI-driven models, especially deep learning algorithms, can predict PD-L1 expression directly from hematoxylin and eosin-stained histological slides, demonstrating high accuracy in estimating PD-L1 expression and predicting responses to immune checkpoint inhibitors across various cancer types [77].

These computational approaches reduce subjectivity associated with manual scoring methods such as Tumor Proportion Score (TPS) and Combined Positive Score (CPS) [77]. Furthermore, integrating AI with multimodal data—including genomics, radiomics, and real-world clinical data—can enhance predictive accuracy and improve patient stratification for immunotherapy [77].

Color Normalization Techniques

Color variation in IHC-stained images, caused by differences in stain operator protocols, exposure times, and slide scanner specifications, significantly impacts feature extraction and interpretation [90]. A novel color normalization technique based on sparse stain separation and self-sparse fuzzy clustering has been developed specifically for breast cancer IHC-stained images [90].

Methodology and Validation:

• Quality Metric: Quaternion structural similarity used to measure normalization algorithm quality
• Validation Technique: Automated and unsupervised nuclei classification with Automatic Color Deconvolution (ACD) tested on normalized images
• Performance: While classification results were similar to other normalization methods, the proposed technique offers easier perception for pathologists

This approach adapts techniques previously used for H&E stained images to IHC staining, despite the difference in perceived colors (two in H&E vs. three in IHC), by developing a structure-preserved normalization method specifically optimized for IHC images [90].

Chromogen Selection and Multiplexing Strategies

The strategic selection of chromogens and development of multiplexing approaches can significantly enhance interpretation of complex staining patterns [91].

Traditional Chromogen Options:

• DAB (3,3'-diaminobenzidine): Robust color, wide dynamic range, highly stable, permanent, insoluble in water and alcohol
• Fast Red-based chromogens: High contrast color but prone to fading and/or blushing when exposed to alcohol or xylene
• Silver Stain: Strong contrast and high sensitivity

Next-Generation Chromogen Technology: Ventana's DISCOVERY chromogens, based on fluorophores to allow unique color generation and narrow-range light absorption, improve compatibility for in situ hybridization (ISH) and IHC multiplexing [91]. These include:

• HRP-driven colors: Purple, Red, Yellow, Blue, Green, Teal
• AP-driven colors: Yellow, Red, Blue

Translucent Chromogens for Co-localization: The availability of translucent chromogens (Purple, Yellow, Teal) enables visualization of overlapping targets in brightfield IHC or ISH multiplexed assays [91]. When biomarkers of interest are present in the same sub-cellular compartment, these chromogens allow color shifts that indicate co-localization:

• Purple and Yellow combine to produce fiery red/orange
• Teal and Purple combine to create indigo blue to deep purple
• Teal and Yellow combine to generate a leafy green color

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Key Research Reagent Solutions for PD-L1 IHC Assay Development

Reagent/Material	Function/Application	Examples/Specifications
Primary Antibodies	Clone-specific binding to PD-L1 epitopes	CAL10, SP263, 22C3, 28-8, SP142 [56] [71]
Staining Platforms	Automated processing of IHC assays	BOND-III (Leica), Benchmark Ultra (Ventana) [56]
Detection Systems	Signal amplification and visualization	HRP or AP-based systems with chromogenic substrates [91] [92]
Chromogens	Produce visible precipitate at antigen sites	DAB (brown), AP Red, DISCOVERY series (Purple, Yellow, Teal) [91]
Counterstains	Provide architectural context and contrast	Hematoxylin (blue nuclei), Nuclear Fast Red (red nuclei), DAPI (fluorescent) [93]
Tissue Controls	Assay validation and quality control	Multi-tissue blocks with tonsil, placenta, known positive/negative samples [56]
Digital Pathology Systems	Whole slide imaging and computational analysis	Aperio GT 450 scanner, digital image analysis algorithms [56] [41]
NIST-Traceable Calibrators	Standardization and harmonization across laboratories	Reference materials with defined unit traceable to SRM 1934 [72]
Color Normalization Tools	Standardize color variation between staining batches	Sparse stain separation, self-sparse fuzzy clustering algorithms [90]

Borderline cases and difficult-to-interpret staining patterns in PD-L1 IHC testing present significant challenges for researchers and drug development professionals. The comparative data and experimental protocols presented in this guide demonstrate that while conventional assays show reasonable concordance, emerging technologies—including quantitative continuous scoring, AI-driven computational pathology, standardized calibration, and advanced chromogen strategies—offer promising avenues for improved objectivity and precision.

The integration of these approaches into standardized laboratory practice requires careful validation and consideration of technical parameters, but holds substantial potential for enhancing patient stratification in immunotherapy. As the field evolves, the adoption of metrologically rigorous tools and computational methods will be crucial for addressing the complexities of PD-L1 expression assessment in borderline cases.

Analytical Validation and Comparative Performance of PD-L1 IHC Assays

The accurate assessment of biomarker expression through Immunohistochemistry (IHC) has become fundamentally important in the era of targeted therapies and immunotherapy, particularly for biomarkers such as PD-L1. In precision oncology, IHC assays directly influence patient selection for specific treatments, making their rigorous validation a critical component of both diagnostic accuracy and drug development. The College of American Pathologists (CAP) guidelines establish a comprehensive framework for analytical validation and ongoing quality assurance of IHC assays to ensure reproducible, reliable, and clinically meaningful results. This comparative performance analysis examines the validation principles and regulatory requirements for IHC assays, with a specific focus on PD-L1 detection, which serves as a cornerstone for immune checkpoint inhibitor therapy. The complex landscape of PD-L1 detection, characterized by multiple antibody clones, different staining platforms, and varying scoring algorithms, creates an imperative for standardized validation approaches that maintain scientific rigor while accommodating necessary methodological diversity [94] [95].

The validation of IHC assays extends beyond technical performance to encompass clinical utility, as evidenced by the direct linkage between PD-L1 expression levels and patient response to immunotherapies such as atezolizumab, durvalumab, and pembrolizumab. Clinical trials have consistently demonstrated that patients with higher PD-L1 expression (as determined by validated IHC assays) derive greater benefit from these treatments, highlighting the critical importance of accurate detection and quantification [96] [95]. Within this context, CAP guidelines provide the foundational principles for establishing assay reliability, while regulatory bodies like the FDA oversee the approval process for companion diagnostics, creating a multi-layered governance structure for IHC assay validation.

Core CAP Guidelines and Regulatory Framework for IHC Validation

Fundamental Validation Principles

The CAP guidelines for IHC assay validation emphasize a systematic approach to establishing analytical performance characteristics before clinical implementation. These guidelines mandate verification that the assay consistently performs according to stated specifications and intended use. While specific CAP guideline documents were not directly available in the search results, the foundational principles referenced in the context of PD-L1 assay development and commercialization include the requirements for specificity, sensitivity, precision, and reproducibility [97] [94]. These analytical parameters form the cornerstone of IHC validation, ensuring that staining patterns accurately reflect true antigen expression rather than technical artifacts.

CAP guidelines further require extensive documentation of pre-analytical, analytical, and post-analytical factors that could impact assay performance. Pre-analytical variables include tissue collection, fixation time, processing methods, and antigen retrieval techniques. Analytical factors encompass antibody clone selection, dilution, incubation times, and detection systems. Post-analytical considerations involve scoring methodologies, pathologist training, and result interpretation criteria. This comprehensive approach acknowledges that robust validation must address the entire testing workflow rather than focusing exclusively on the antibody-antigen interaction [97]. For companion diagnostics, CAP guidelines align with FDA requirements, mandating more stringent validation protocols compared to laboratory-developed tests (LDTs), including defined performance thresholds for sensitivity and specificity against clinical endpoints.

Regulatory Pathways: Companion Diagnostics vs. LDTs

The regulatory landscape for IHC assays operates through two primary pathways: FDA-approved companion diagnostics and Laboratory Developed Tests (LDTs). FDA-approved companion diagnostics, such as the Ventana SP263 assay, undergo the most rigorous validation process, requiring evidence of clinical utility obtained through large-scale clinical trials that directly link test results to therapeutic response [97] [94]. These assays must demonstrate robust reproducibility across multiple laboratories and staining platforms, with predefined scoring criteria that maintain consistency in interpretation.

In contrast, LDTs validated according to CAP guidelines implement a similarly structured analytical validation framework but may lack the extensive clinical outcome data required for FDA approval. The 2024 FDA laboratory-developed test rules are expected to significantly impact this landscape, potentially requiring between $566 million and $3.56 billion in compliance expenditures across the industry according to market analyses [97]. This regulatory evolution reflects increasing recognition of the critical role that standardized biomarker detection plays in patient care, particularly for biomarkers like PD-L1 where expression thresholds directly influence treatment decisions.

Comparative Performance of PD-L1 IHC Assays

Antibody Clone Comparison and Analytical Performance

The performance of PD-L1 IHC assays varies considerably depending on the antibody clone utilized, with different clones exhibiting distinct staining characteristics and interpretation criteria. Recent developments have focused on identifying cost-effective alternatives to commercially approved assays without compromising analytical performance.

Table 1: Comparative Performance of PD-L1 Antibody Clones in Lung Adenocarcinoma

Antibody Clone	Comparator Clone	Overall Accuracy	Kappa Value	Key Characteristics
3E2 (Novel)	28-8	90.1%	0.797	Cost-effective alternative; potential for 30-60% cost savings
3E2 (Novel)	E1L3N	69.8%	0.401	Moderate agreement
3E2 (Novel)	SP263	55.4%	0.262	Low agreement
SP263 (FDA-approved)	N/A	Reference standard	N/A	High reliability but with significant cost barriers

A 2025 study evaluating the novel 3E2 clone demonstrated its strong agreement with the established 28-8 clone, with an overall accuracy of 90.1% and a kappa value of 0.797, indicating high concordance [94]. This performance profile suggests that the 3E2 clone may represent a viable alternative in resource-constrained settings, potentially expanding access to PD-L1 testing. However, the same study revealed substantially lower agreement with other clones, particularly the SP263 assay, highlighting the significant variability that can exist between different antibody clones despite targeting the same biomarker [94].

Clinical Performance and Predictive Value

Beyond analytical performance, the clinical utility of PD-L1 assays is ultimately determined by their ability to predict treatment response to immune checkpoint inhibitors. Clinical trials have established consistent correlations between PD-L1 expression levels and therapeutic outcomes across multiple cancer types.

Table 2: Clinical Performance of PD-L1 Assays in Predicting Immunotherapy Response

Clinical Trial	Therapeutic Agent	PD-L1 Threshold	Objective Response Rate	Overall Survival Benefit
BIRCH	Atezolizumab	TC3 or IC3	34%	Median OS: 23.5 months
BIRCH	Atezolizumab	TC2 or IC2	18%	Similar OS trend
ATLANTIC	Durvalumab	≥90%	30.9%	1-year OS: 50.8%
ATLANTIC	Durvalumab	≥25%	16.4%	1-year OS: 47.7%
ATLANTIC	Durvalumab	<25%	7.5%	1-year OS: 34.5%
CheckMate 057	Nivolumab	PD-L1 positive	19%	Median OS: 12.2 months

The BIRCH trial demonstrated that patients with the highest PD-L1 expression (TC3 or IC3) achieved an objective response rate of 34% with atezolizumab monotherapy, compared to 18% for those with intermediate expression (TC2 or IC2) [96]. Similarly, the ATLANTIC trial showed a clear gradient of response based on PD-L1 expression levels, with the highest expression cohort (≥90%) achieving an objective response rate of 30.9%, compared to just 7.5% in the low/negative expression group [96]. These findings underscore the critical importance of accurate PD-L1 quantification in identifying patients most likely to benefit from immunotherapy, thereby validating the clinical utility of properly validated IHC assays.

Experimental Protocols for IHC Assay Validation

Methodologies for Comparative Antibody Performance Studies

Robust experimental protocols are essential for meaningful comparison of IHC assay performance. The evaluation of the novel 3E2 PD-L1 antibody clone followed a rigorous methodology that can serve as a template for comparative validation studies [94]:

Sample Cohort Selection: The study utilized 101 formalin-fixed, paraffin-embedded (FFPE) lung adenocarcinoma tissue samples obtained from surgical resections or biopsies between May 2018 and November 2021. Key inclusion criteria included pathological confirmation of lung adenocarcinoma and absence of preoperative chemotherapy, radiotherapy, or targeted therapy, which could potentially alter PD-L1 expression patterns.

Immunohistochemical Staining Protocol: The experimental workflow involved parallel staining of consecutive tissue sections from the same patient blocks using four different antibody clones: the experimental 3E2 clone, Abcam 28-8, CST E1L3N, and Ventana SP263. This direct comparison on adjacent sections from the same tissue blocks minimized pre-analytical variables and enabled direct comparison of staining performance. The protocol included standardized antigen retrieval conditions, antibody incubation times, and detection systems to ensure methodological consistency across compared clones.

Quantitative Analysis and Concordance Assessment: Stained slides were digitally scanned and quantitatively analyzed using image analysis software. Statistical measures included Bland-Altman plots for assessing agreement between continuous measurements, calculation of overall accuracy, and Cohen's kappa coefficient for categorical concordance. The kappa statistic is particularly important in validation studies as it measures inter-rater agreement beyond what would be expected by chance alone, with values above 0.75 generally considered excellent agreement [94].

Clinical Correlation: To establish clinical validity, the study included survival analysis of 16 stage III-IV lung adenocarcinoma patients who received immunotherapy, correlating PD-L1 expression levels (as determined by the 3E2 clone) with overall survival. This critical step connects analytical performance to clinical outcomes, demonstrating that PD-L1 expression ≥5% as detected by 3E2 was associated with significantly better survival (p=0.021), mirroring results obtained with the established 28-8 clone (p=0.019) [94].

Emerging Methodologies: Multiplex Immunohistochemistry

While conventional IHC detects single biomarkers, multiplex IHC represents an advanced methodology enabling simultaneous detection of multiple biomarkers on a single tissue section. This technology employs tyramide signal amplification (TSA) and antibody stripping techniques to sequentially label multiple antigens without cross-reactivity [98]. The validation of multiplex IHC assays requires additional considerations beyond conventional IHC, including:

Validation of Antibody Stripping Efficiency: Protocols must demonstrate complete removal of primary and secondary antibodies between staining rounds without damaging subsequent epitopes or causing tissue loss.

Spectral Unmixing Validation: For fluorescent multiplex IHC, rigorous validation of spectral unmixing algorithms is essential to ensure that signals from different fluorophores are accurately distinguished without bleed-through or crossover.

Spatial Analysis Validation: Multiplex IHC enables spatial analysis of cellular interactions within the tumor microenvironment. Validation of spatial analysis algorithms requires demonstration of accuracy in cell identification, segmentation, and spatial relationship quantification.

The implementation of multiplex IHC offers significant advantages for tumor microenvironment analysis, including preserved spatial relationships between different cell types, reduced tissue consumption, and comprehensive immunoprofiling from limited sample material [98]. However, these benefits come with increased validation complexity, particularly regarding antibody panel optimization and computational analysis validation.

IHC Assay Validation Workflow and Key Components

The Scientist's Toolkit: Essential Reagents and Materials

The validation and implementation of robust IHC assays for PD-L1 detection requires specific reagents and materials that ensure reproducibility, accuracy, and compliance with regulatory standards. The following toolkit encompasses essential components referenced in the search results:

Table 3: Essential Research Reagent Solutions for IHC Assay Validation

Reagent/Material	Function	Examples/Characteristics
Primary Antibody Clones	PD-L1 Epitope Binding	3E2, 28-8, SP263, 22C3; Specific clone selection affects staining intensity and pattern
Detection System	Signal Amplification	Polymer-based systems; HRP-conjugated secondaries; TSA amplification for multiplex IHC
Automated Staining Platforms	Standardization	Ventana Benchmark ULTRA; Leica Bond RX; Enable batch processing and protocol consistency
Antigen Retrieval Reagents	Epitope Exposure	Citrate-based (pH 6.0) or EDTA/TRIS-based (pH 9.0) buffers; Optimization required for each antibody
Positive Control Tissues	Assay Validation	Placental tissue; tonsil; Known PD-L1 expressing cell lines; Essential for daily run validation
Digital Pathology Systems	Quantification and Analysis	Slide scanners with image analysis software; Enable standardized scoring and algorithm deployment
Multiplex IHC Reagents	Simultaneous Multi-Marker Detection	Celnovte mIHC kits; Opal systems; Allow TSA-based sequential staining with antibody stripping

The selection of appropriate primary antibody clones represents perhaps the most critical decision in IHC assay development. Studies comparing the novel 3E2 clone with established assays demonstrated that while cost-effective alternatives exist, thorough validation against reference standards is essential [94]. The growing adoption of multiplex IHC reagents, such as those in the Celnovte product line, reflects the increasing demand for simultaneous evaluation of multiple biomarkers within the spatial context of the tumor microenvironment [98]. These systems typically include enzyme-labeled polymers, TSA dyes across various fluorescence channels (e.g., CM480 to CM780 series), and DAPI nuclear counterstains to facilitate comprehensive tissue analysis.

For automated staining platforms, integration of pre-optimized reagent kits that minimize protocol adjustments enhances reproducibility across laboratories. The consistent performance of these systems depends heavily on standardized antigen retrieval methods and calibrated detection chemistry that maintain lot-to-lot consistency—a particular challenge noted in the search results, which indicated that approximately two-thirds of commercial antibodies fail basic specificity testing, forcing laboratories to implement costly internal validation procedures [97]. This underscores the importance of obtaining reagents from manufacturers with robust quality control systems, such as those with GMP manufacturing conditions and ISO9001/ISO13485 certifications [98].

The validation of IHC assays for PD-L1 detection represents a dynamic interface between diagnostic pathology, regulatory science, and clinical oncology. CAP guidelines and regulatory requirements establish essential frameworks for ensuring assay reliability, but practical implementation requires careful navigation of technical and economic challenges. The comparative analysis presented herein demonstrates that while cost-effective alternatives to commercial assays continue to emerge—such as the 3E2 clone with its potential for 30-60% cost savings—their adoption must be guided by rigorous validation against both analytical standards and clinical outcomes [94].

The evolving landscape of IHC validation is increasingly shaped by advanced methodologies such as multiplex staining and AI-assisted workflows that enhance quantification objectivity and reproducibility. The integration of digital pathology platforms with cloud-based AI algorithms, as exemplified by systems like Roche's navify digital pathology, creates opportunities for standardized analysis while simultaneously introducing new validation considerations for computational components [97]. Furthermore, the growing emphasis on companion diagnostic approvals continues to raise the validation bar, requiring increasingly robust evidence of clinical utility tied directly to therapeutic response [97] [95].

As immunotherapy treatment paradigms expand across cancer types, the principles of IHC assay validation will continue to evolve in complexity and importance. The fundamental requirement remains unchanged: ensuring that diagnostic assays provide reproducible, accurate, and clinically meaningful results that optimally guide patient care. Through adherence to structured validation frameworks, implementation of comprehensive experimental protocols, and utilization of quality-controlled reagents, laboratories can navigate this challenging landscape while advancing the field of precision cancer diagnostics.

The advent of immune checkpoint inhibitors (ICIs) has revolutionized the treatment landscape for advanced cancers, notably non-small cell lung cancer (NSCLC). Programmed death-ligand 1 (PD-L1) expression on tumor cells, as detected by immunohistochemistry (IHC), serves as a primary predictive biomarker for patient selection in anti-PD-1/PD-L1 therapies [6] [71]. However, a significant challenge has emerged in clinical practice: for each distinct ICI, a unique, corresponding PD-L1 IHC assay was developed and validated within its specific clinical trial [71]. This has resulted in a proliferation of companion and complementary diagnostics (e.g., the Dako 22C3 and 28-8 assays, and the Ventana SP263 and SP142 assays), each with its own protocol, scoring algorithm, and approved clinical purpose [6].

This multiplicity of tests is not practically or economically feasible for most pathology laboratories to implement simultaneously, given constraints of cost, tissue availability, and staining platforms [71]. Consequently, the central question of whether these various PD-L1 IHC assays are "interchangeable" has become a critical area of investigation. This guide objectively compares the performance of standardized PD-L1 assays and laboratory-developed tests (LDTs), framing the discussion within the broader thesis of comparative performance and diagnostic accuracy for PD-L1 detection. It synthesizes evidence from key meta-analyses and systematic reviews to provide researchers, scientists, and drug development professionals with a data-driven resource for understanding assay concordance and its implications for clinical practice and biomarker development.

Data from large-scale meta-analyses and systematic reviews provide the most robust evidence regarding assay comparability. The findings are summarized in the tables below.

Table 1: Key Findings from Meta-Analyses and Systematic Reviews on PD-L1 Assay Interchangeability

Study Focus	Included Studies & Samples	Key Findings on Interassay Concordance	Key Findings on Interobserver Concordance	Conclusion on Interchangeability
Meta-Analysis of Diagnostic Accuracy [6]	22 studies; 376 assay comparisons; primarily NSCLC.	For specific clinical purposes (e.g., TPS ≥1% or ≥50%), replacing an FDA-approved CDx with another CDx developed for a different purpose often resulted in diagnostic sensitivity/specificity <90%. Properly validated LDTs could be a better alternative.	Not the primary focus.	Assays are not automatically interchangeable for a purpose they were not clinically validated for. A purpose-based approach is essential.
Systematic Review of Comparability [71]	27 studies; sample sizes 15-713 NSCLC specimens.	High analytical concordance between 22C3, 28-8, and SP263 assays, and properly validated LDTs. Lower concordance observed in comparisons involving the SP142 assay. Concordance decreased when using clinical cut-offs.	High interobserver agreement for all assays/LDTs, but lower agreement at the 1% cut-off compared to the 50% cut-off.	Interchangeability is hampered by the use of cut-offs. Discordance at the 1% cut-off may deny patients treatment.

Table 2: Comparative Performance of Different PD-L1 Assay Clones

Assay Clone	Associated Drug(s)	Comparability with 22C3, 28-8, SP263	Key Limitations & Considerations
22C3	Pembrolizumab	High overall agreement [71].	FDA-approved as a companion diagnostic. Interchangeability with others is context-dependent [6].
28-8	Nivolumab	High overall agreement [71].	FDA-approved as a complementary diagnostic.
SP263	Durvalumab	High overall agreement [71].	CE-marked for use with pembrolizumab and nivolumab, indicating recognized similarity [71].
SP142	Atezolizumab	Lower concordance consistently observed [71].	Tends to report lower Tumor Cell (TC) staining percentages, leading to more false negatives if used interchangeably [71].

Table 3: Pathologist vs. Artificial Intelligence (AI) Performance in PD-L1 Scoring [9]

Scoring Method	Interobserver Agreement (TPS ≥50%)	Interobserver Agreement (TPS <1%)	Intraobserver Consistency	Agreement with Median Pathologist Score (TPS ≥50%)
Pathologists (Light Microscopy/WSI)	Almost perfect (Fleiss' kappa = 0.873)	Moderate (Fleiss' kappa = 0.558)	High (Cohen's kappa: 0.726 - 1.0)	(Reference Standard)
AI Algorithm (uPath - Roche)	Not Reported	Not Reported	Not Reported	Fair (Fleiss' kappa = 0.354)
AI Algorithm (Visiopharm)	Not Reported	Not Reported	Not Reported	Substantial (Fleiss' kappa = 0.672)

Experimental Protocols in Cited Studies

The conclusions drawn in the cited meta-analyses and reviews rely on rigorous experimental and methodological protocols. This section details the core methodologies employed by the primary studies that contributed to these aggregated findings.

Systematic Literature Review and Meta-Analysis Protocol

The meta-analysis by Munari et al. (2020) followed a structured, pre-defined protocol [6]:

• Literature Search: A systematic search of the MEDLINE database (via PubMed) was conducted for the period January 2015 to August 2018 using the term "PD-L1", limited to English-language human studies.
• Study Selection: From 2,515 identified abstracts, 57 full-text articles on comparisons of two or more PD-L1 assays were reviewed. Ultimately, 22 publications were selected for meta-analysis.
• Data Abstraction and Quality Assessment: Modified GRADE and QUADAS-2 criteria were used to grade published evidence and design data abstraction templates. Additional raw data (e.g., 2x2 contingency tables) were requested from authors of 20 of the 22 studies to enable the meta-analysis.
• Statistical Analysis: Data were pooled using a random-effects model. The primary outcome measures were diagnostic sensitivity and specificity for specific clinical purposes (e.g., TPS ≥1% or ≥50%). Assays were considered clinically acceptable if both sensitivity and specificity were ≥90%.

Primary Study Design for Assay Comparison

The individual studies included in the systematic reviews, such as the Blueprint project, typically employed the following experimental workflow [71]:

• Sample Cohort: Selection of a set of archived NSCLC tissue samples (e.g., surgical resections and biopsies), often formatted into tissue microarrays (TMAs).
• Staining Protocol: Consecutive tissue sections from each sample are stained with different PD-L1 IHC assays (e.g., 22C3, 28-8, SP263, SP142) and/or LDTs, following the manufacturers' instructions for use on their respective staining platforms (Dako or Ventana).
• Scoring by Pathologists: The stained slides are scored independently by multiple pathologists who are typically blinded to the assay type and other pathologists' scores. The scoring is based on Tumor Proportion Score (TPS), which is the percentage of viable tumor cells showing partial or complete membranous staining at any intensity.
• Statistical Analysis for Concordance: Concordance is evaluated using statistical measures such as the Pearson's or Spearman's correlation coefficient for continuous scores, and the Cohen's kappa or Intraclass Correlation Coefficient (ICC) for agreement at specific cut-offs (1%, 50%). Overall Percentage Agreement (OPA), Positive Percentage Agreement (PPA), and Negative Percentage Agreement (NPA) are also commonly calculated.

Protocol for Evaluating Digital Pathology and AI

The study by Leithner et al. (2025) provides a template for comparing pathologist and AI performance [9]:

• Sample and Staining: 51 NSCLC cases were stained with the VENTANA PD-L1 (SP263) assay.
• Pathologist Scoring: Six pathologists scored each case twice: first via light microscopy, and after a washout period of at least one month, via whole-slide images (WSI). Scores were recorded in specific increments.
• AI Scoring: Two commercially available AI algorithms (Roche's uPath and Visiopharm's PD-L1 Lung Cancer TME application) were used to score the same WSIs.
• Analysis of Agreement: Fleiss' kappa was used to assess interobserver agreement among pathologists. Cohen's kappa was used for intraobserver consistency. Agreement between AI algorithms and the median pathologist score was also calculated using Fleiss' kappa.

Visualizing the Experimental Workflow and Decision Pathway

The following diagrams, generated using Graphviz, illustrate the core experimental workflow for assay comparison studies and the logical decision pathway for determining assay interchangeability in a clinical context.

Experimental Workflow for PD-L1 Assay Comparison Studies

Decision Pathway for PD-L1 Assay Interchangeability

The Scientist's Toolkit: Key Research Reagent Solutions

The following table details essential materials and reagents used in PD-L1 IHC testing and comparative research.

Table 4: Essential Reagents and Materials for PD-L1 IHC Research

Item Name	Function/Description	Example Specifics
FDA-Approved Companion Diagnostic Kits	Standardized, regulatory-cleared tests for determining patient eligibility for specific drugs.	PD-L1 IHC 22C3 PharmDx (Dako/Agilent) for pembrolizumab; PD-L1 IHC 28-8 PharmDx (Dako/Agilent) for nivolumab [71].
Ventana PD-L1 (SP263) Assay	A standardized assay widely used in comparison studies and approved for use with durvalumab.	Often shows high concordance with 22C3 and 28-8 [71]. Used on Ventana BenchMark ULTRA platform [9].
Laboratory-Develop Test (LDT) Antibodies	In-house validated tests using commercially available antibodies, offering potential flexibility and cost savings.	Clones like E1L3N, SP142, etc. Must be properly validated against a reference standard for a specific clinical purpose [6] [71].
Automated IHC Staining Platforms	Instruments that automate the staining process to improve reproducibility and reduce variability.	Dako Omnis or Link 48 (for Dako assays); Ventana BenchMark ULTRA or XT (for Ventana assays) [71]. Platform differences can contribute to variability.
Whole-Slide Scanners & Digital Pathology	Hardware for digitizing glass slides, enabling remote review, archival, and use with AI algorithms.	Scanners like the Ventana DP200 (Roche) or PANORAMIC1000 (3DHISTECH) are used to create whole-slide images for pathologist and AI review [9].
AI-Powered Image Analysis Software	Algorithms designed to automatically detect tumor regions and quantify PD-L1 TPS, reducing subjectivity.	Commercial applications include Roche uPath and Visiopharm PD-L1 Lung Cancer TME [9]. Performance compared to pathologists is an active research area.

Immunohistochemistry (IHC) assays for programmed death-ligand 1 (PD-L1) expression serve as critical companion diagnostics for immune checkpoint inhibitor therapies across multiple malignancies. The comparative performance of these assays varies significantly across different tumor types due to biological differences in PD-L1 expression patterns, scoring algorithms, and tissue microenvironment characteristics. This review systematically evaluates the technical performance of major PD-L1 IHC assays—including the Dako 22C3, Ventana SP263, Ventana SP142, and emerging assays—across three major cancer types: non-small cell lung cancer (NSCLC), head and neck squamous cell carcinoma (HNSCC), and urothelial carcinoma (UC). Understanding these assay-specific performance characteristics is essential for appropriate test selection in both clinical trials and routine practice, ensuring accurate patient stratification for immunotherapy.

Comparative Performance Data Across Tumor Types

Assay Concordance in Non-Small Cell Lung Cancer (NSCLC)

In NSCLC, multiple studies have demonstrated strong concordance between the 22C3, SP263, and 28-8 assays, while the SP142 assay consistently shows lower sensitivity for tumor cell staining.

Table 1: PD-L1 Assay Concordance in NSCLC (Tumor Cell Scoring)

Compared Assays	Statistical Measure	1% Cut-off	25% Cut-off	50% Cut-off	Study
SP263 vs 22C3	Kappa (κ)	0.71	0.75	0.81	Ring Study [99]
SP263 vs 22C3	Overall Percent Agreement	-	-	Lower bound 95% CI: 86.2%	CAL10 Development Study [56]
Various Assays & LDTs	Systematic Review Conclusion	High agreement between 22C3, 28-8, SP263; Lower concordance with SP142	-	-	Systematic Review [71]

The Ring Study, an international comparison conducted across multiple centers, demonstrated almost perfect agreement between SP263 and 22C3 at the 50% cut-off (κ=0.81), which was superior to the agreement at lower cut-offs [99]. This high level of concordance supports potential interchangeability between these assays in NSCLC, particularly at the clinically relevant 50% threshold for pembrolizumab monotherapy. A recent developmental study for the novel CAL10 assay demonstrated comparable performance to the SP263 assay, with the lower bound of the 95% confidence interval for overall percent agreement reaching 86.2% at the ≥50% tumor proportion score (TPS) cut-off [56].

Performance in Head and Neck Squamous Cell Carcinoma (HNSCC)

Table 2: PD-L1 Assay Performance in HNSCC

Performance Aspect	Findings	Assay	Study
Interobserver Agreement (25% cut-off)	κ=0.60 to 0.82	SP263	Ring Study [99]
Interobserver Agreement (50% cut-off)	κ=0.64 to 0.90	SP263	Ring Study [99]
73-10 Clone Positivity Rate	79% in HNSCC vs 3% in normal mucosa	73-10	High-Sensitivity Study [100]
Specimen Type Discrepancies	Significant differences in CPS/TPS between biopsy and resection (p<0.01)	22C3	HNSCC Specimen Comparison [57]

The Ring Study demonstrated that the performance of the SP263 assay in HNSCC is comparable across five different countries, indicating robust international consistency [99]. Recent research has highlighted the impact of specimen type on PD-L1 scoring in HNSCC. A comprehensive study of 68 HNSCC cases found significant discrepancies in both combined positive score (CPS) and TPS between preoperative biopsy and surgical resection specimens (p<0.01), as well as between surgical resection and metastatic lymph nodes (p<0.01) [57]. This heterogeneity emphasizes the challenges in PD-L1 assessment in HNSCC and suggests that specimen type must be considered when interpreting results.

Emerging data on the high-sensitivity 73-10 clone demonstrates its potential utility in HNSCC, with a 79% positivity rate (using tumor cell score ≥1%) in HNSCC compared to only 3% in normal oral mucosa [100]. This clone also correlated with high CD4+ tumor-infiltrating lymphocytes and served as an independent prognostic factor for overall survival, disease-specific survival, and progression-free survival.

Challenges in Urothelial Carcinoma

Urothelial carcinoma presents unique challenges for PD-L1 assessment due to the importance of immune cell staining and considerable discordance between assays.

Table 3: PD-L1 Assay Performance in Urothelial Carcinoma

Performance Aspect	Findings	Assays Compared	Study
Interobserver Agreement (TC, 25% cut-off)	κ=0.68 to 0.91	SP263	Ring Study [99]
Immune Cell Scoring Concordance	Poor to substantial (κ= -0.04 to 0.76)	SP263	Ring Study [99]
Immune Cell Scoring Correlation	Low (CCC=0.10 to 0.68)	SP263	Ring Study [99]
Biological Basis of Discordance	SP142 preferentially detects PD-L1 on dendritic cells	SP142 vs 22C3	IMvigor130 Analysis [101]
Clinical Outcome Association	"22C3-only positive" associated with worse outcomes	SP142 vs 22C3	IMvigor130 Analysis [101]

The Ring Study highlighted particular challenges in UC, with low concordance for immune cell staining across different cut-offs (1%, 5%, 10%, and 25%), which may significantly impact treatment decisions [99]. A detailed analysis of specimens from the IMvigor130 trial revealed that discordance between the SP142 and 22C3 assays stems from their detection of biologically distinct PD-L1 expression patterns [101]. The SP142 assay preferentially detects PD-L1-expressing dendritic cells, which are associated with more favorable outcomes to immune checkpoint blockade, while cases positive only by the 22C3 assay (associated with tumor cell-dominant PD-L1 expression) correlated with worse outcomes.

A smaller comparative study of 24 UC cases found generally good concordance among the three antibodies (SP142, SP263, and 22C3), though it noted consistent underestimation of PD-L1 expression by the SP142 clone compared to the others [102].

Experimental Methodologies in Key Studies

The Ring Study Protocol

The Ring Study implemented a rigorous multicenter design to assess PD-L1 assay performance across NSCLC, HNSCC, and UC [99]. Excisional specimens from each cancer type were assayed using the Ventana SP263 platform at three sites in six countries (Australia, Brazil, Korea, Mexico, Russia, and Taiwan). All stained slides were rotated to two other sites for interobserver scoring. In the NSCLC cohort, the same tissue samples were also assessed with the Dako 22C3 pharmDx assay for direct comparison. PD-L1 immunopositivity was scored according to approved algorithms: the percentage of PD-L1-expressing tumor cells for SP263 and tumor proportion score for 22C3. Statistical analysis included kappa statistics for categorical agreement and concordance correlation coefficients for continuous measures.

Novel Assay Development Methodology

The developmental study for the Leica Biosystems PD-L1 CAL10 assay employed a feasibility design comparing the novel assay to the established SP263 assay in NSCLC samples [56]. The study included 136 formalin-fixed paraffin-embedded NSCLC tissue samples with case characteristics reflecting real-world diversity: 76 resection specimens, 23 biopsies, 88 adenocarcinomas, 43 squamous cell carcinomas, and one large cell carcinoma. Cases were pre-screened and pre-characterized using the BOND RTU PD-L1 (73-10) clone to ensure representation across the full TPS range (0-100%). Staining was performed on the BOND-III system for CAL10 and the Benchmark Ultra system for SP263, with appropriate controls. Two pathologists independently read randomized, anonymized slide sets, with statistical analysis focusing on overall percent agreement with predefined non-inferiority targets.

Digital Pathology and Computational Approaches

Advanced computational methods for PD-L1 assessment are emerging as alternatives to visual scoring. The PD-L1 Quantitative Continuous Scoring (QCS) system utilizes computer vision for granular cell-level quantification of PD-L1 staining intensity in digitized whole slide images [41]. This approach derives a biomarker capturing the percentage of tumor cells with medium to strong staining intensity (PD-L1 QCS-PMSTC), classifying patients with ≥0.575% as biomarker-positive. When validated in the MYSTIC trial (768 whole slide images), this quantitative method achieved a hazard ratio of 0.62 (CI 0.46-0.82) for durvalumab versus chemotherapy, with a substantially increased biomarker-positive prevalence of 54.3% compared to 29.7% for visual TPS ≥50% scoring.

Another study employed open-source bioimage analysis using QuPath software to manually annotate four distinct cell populations: tumor cells, immune cells, PD-L1-expressing tumor cells, and PD-L1-expressing immune cells in HNSCC samples [57]. This digital approach enabled precise quantification of CPS and TPS across 204 tissue sections from 68 patients, revealing significant differences between specimen types that might be challenging to detect with visual scoring alone.

Analytical Framework and Technical Considerations

PD-1/PD-L1 Signaling Pathway

The PD-1/PD-L1 pathway represents a critical immune checkpoint mechanism that regulates T-cell-mediated immunity. Under normal physiological conditions, the interaction between PD-L1 on antigen-presenting cells and PD-1 on T-cells maintains immune tolerance and prevents excessive inflammation [56]. Tumor cells exploit this mechanism by overexpressing PD-L1, which engages PD-1 on tumor-infiltrating T-cells, leading to T-cell exhaustion and immune evasion [56] [100]. Immune checkpoint inhibitors, including anti-PD-1 (pembrolizumab, nivolumab) and anti-PD-L1 (atezolizumab, durvalumab) antibodies, block this interaction, restoring T-cell-mediated anti-tumor immunity [56] [101]. The accurate assessment of PD-L1 expression through IHC assays is therefore essential for identifying patients most likely to benefit from these therapies.

Comparative PD-L1 IHC Testing Workflow

The standardized workflow for comparative PD-L1 IHC assay evaluation begins with formalin-fixed paraffin-embedded tissue blocks sectioned at 4µm thickness [57] [100]. Sections undergo automated IHC staining on dedicated platforms (Dako, Ventana, or Leica systems) using validated protocols for each antibody clone [56] [71]. Following staining, slides are digitized using whole slide scanners (Aperio GT 450, Philips Intellisite, or PANNORAMIC systems) to enable both visual assessment by pathologists and computational analysis [56] [57] [41]. Evaluation incorporates multiple scoring systems: Tumor Proportion Score (percentage of PD-L1-positive tumor cells), Combined Positive Score (number of PD-L1-positive cells divided by total tumor cells multiplied by 100), and Immune Cell scoring (percentage of tumor area occupied by PD-L1-positive immune cells) [57] [101]. Statistical analysis focuses on concordance metrics (kappa statistics, overall percent agreement, concordance correlation coefficients), interobserver variability, and correlation with clinical outcomes [99] [71].

Research Reagent Solutions

Table 4: Key Reagents and Platforms for PD-L1 IHC Research

Category	Specific Product	Research Application	Performance Notes
Commercial Assays	PD-L1 IHC 22C3 pharmDx (Dako)	Companion diagnostic for pembrolizumab	High concordance with SP263 in NSCLC [99] [71]
	PD-L1 IHC SP263 (Ventana)	Companion diagnostic for durvalumab	Comparable performance across tumor types [99]
	PD-L1 IHC SP142 (Ventana)	Complementary diagnostic for atezolizumab	Lower tumor cell sensitivity; preferential dendritic cell staining [101] [71]
Development Assays	PD-L1 CAL10 (Leica)	Novel assay in development	Comparable to SP263 in NSCLC [56]
	PD-L1 73-10 (Leica)	High-sensitivity detection	Superior sensitivity in HNSCC [100]
Staining Platforms	Dako Autostainer Link	22C3 assay platform	-
	Ventana Benchmark Ultra	SP263, SP142 assay platform	-
	Leica BOND-III	CAL10, 73-10 assay platform	-
Digital Analysis Tools	QuPath	Open-source bioimage analysis	Cell classification and scoring [57]
	HALO	Commercial image analysis	Multiplex analysis and co-localization [101]
	Aperio GT 450	Whole slide imaging	Digital slide creation [56]

The comparative performance of PD-L1 IHC assays varies significantly across NSCLC, HNSCC, and urothelial carcinoma, reflecting biological differences in PD-L1 expression patterns and tumor microenvironment characteristics. In NSCLC, the 22C3, SP263, and 28-8 assays demonstrate strong concordance, supporting potential interchangeability with proper validation, while the SP142 assay shows consistently lower tumor cell sensitivity. HNSCC displays significant PD-L1 heterogeneity across different specimen types, complicating clinical assessment. Urothelial carcinoma presents unique challenges due to biologically meaningful discordance between assays, with the SP142 assay preferentially detecting dendritic cell PD-L1 expression associated with better outcomes. Emerging technologies including high-sensitivity clones (73-10, CAL10), digital pathology platforms, and quantitative continuous scoring systems show promise for improving the accuracy and reproducibility of PD-L1 assessment. Optimal PD-L1 testing requires careful consideration of tumor type, specimen characteristics, scoring systems, and clinical context to ensure appropriate patient selection for immunotherapy.

The assessment of Programmed Death-Ligand 1 (PD-L1) expression via immunohistochemistry (IHC) has become a critical predictive biomarker in oncology, guiding treatment decisions for immune checkpoint inhibitors across multiple cancer types including non-small cell lung cancer (NSCLC), gastric cancer, and bladder cancer [103] [58] [104]. The prevailing clinical standard involves visual scoring by pathologists, primarily using the Tumor Proportion Score (TPS) or Combined Positive Score (CPS). However, this manual approach suffers from significant interobserver variability and subjectivity [105] [41]. The emergence of artificial intelligence (AI) algorithms for digital pathology promises to enhance the accuracy, standardization, and efficiency of PD-L1 scoring. This article provides a comprehensive comparison of pathologist-based and AI-based PD-L1 scoring methodologies, examining their respective performance characteristics, technical approaches, and implications for clinical practice and drug development.

Performance Comparison: Pathologists vs. AI Algorithms

Concordance and Variability Metrics

Table 1: Comparative Performance of Pathologists and AI Algorithms in PD-L1 Scoring

Assessment Method	Cancer Type	Metric	Performance Value	Reference Standard
Pathologists (Interobserver)	NSCLC	Fleiss' Kappa (TPS ≥50%)	0.873 (almost perfect)	Median pathologist scores [103]
Pathologists (Interobserver)	NSCLC	Fleiss' Kappa (TPS <1%)	0.558 (moderate)	Median pathologist scores [103]
Pathologists (Intraobserver)	NSCLC	Cohen's Kappa Range	0.726 - 1.0	Self-consistency [103]
Visiopharm AI Algorithm	NSCLC	Fleiss' Kappa (TPS ≥50%)	0.672 (substantial)	Median pathologist scores [103]
uPath (Roche) AI Algorithm	NSCLC	Fleiss' Kappa (TPS ≥50%)	0.354 (fair)	Median pathologist scores [103]
Deep Learning Model (Lunit)	NSCLC	Spearman Correlation	0.925	Pathologist consensus [106]
YOLO-based AI Pipeline	Gastric Cancer	Cohen's Kappa (CPS)	0.782	Expert pathologist consensus [58]
QuPath (AI-SAI Protocol)	Bladder Cancer	Cohen's Kappa	0.86	Manual assessment [104]
QuPath (AI-WSI Protocol)	Bladder Cancer	Cohen's Kappa	0.65	Manual assessment [104]

Clinical Predictive Value

Table 2: Predictive Performance for Immunotherapy Response

Scoring Method	Cancer Type	Predictive Metric	Performance	Clinical Context
Pathologist TPS (≥50%)	NSCLC (MYSTIC Trial)	Hazard Ratio (PFS)	0.69 (CI 0.46-1.02)	Durvalumab vs Chemotherapy [41]
PD-L1 QCS-PMSTC (>0.575%)	NSCLC (MYSTIC Trial)	Hazard Ratio (PFS)	0.62 (CI 0.46-0.82)	Durvalumab vs Chemotherapy [41]
AI-Powered TPS (Lunit)	NSCLC	Hazard Ratio (PFS, TPS <1%)	2.38 (CI 1.69-3.35)	ICI treatment [106]
Pathologist TPS	NSCLC	Hazard Ratio (PFS, TPS <1%)	1.62 (CI 1.23-2.13)	ICI treatment [106]
AI Spatial Biomarker	NSCLC	Hazard Ratio (PFS)	5.46	ICI treatment [107]
PD-L1 TPS Alone	NSCLC	Hazard Ratio (PFS)	1.67	ICI treatment [107]

Experimental Protocols and Methodologies

Pathologist Scoring Protocols

Traditional pathologist assessment of PD-L1 expression typically follows standardized protocols. In a comparative study involving 51 SP263-stained NSCLC cases, six pathologists (five pulmonary pathologists and one in training) scored slides using both light microscopy and whole-slide images (WSI) [103] [9]. The evaluation was performed with a washout period of at least one month between light microscopy and digital scoring to minimize recall bias, following CAP-PLQC guidelines [9]. Pathologists evaluated only tumor cells, considering any intensity of either partial or complete membranous staining as positive. The percentage of positively stained tumor cells was recorded categorically: 0%, 1%, 5%, 10%, and up to 100% in 10% increments [9]. This methodology reflects real-world clinical practice where pathologists visually estimate the proportion of positive tumor cells relative to all viable tumor cells.

AI Algorithm Training and Implementation

AI approaches demonstrate considerable diversity in their technical implementation. Most systems employ a multi-stage pipeline combining computer vision and deep learning architectures:

Dual-Network Architecture for Tumor Region Identification: The gastric cancer CPS system developed by [58] employs a pipeline that first identifies tumor regions using a combination of MobileNet for patch-level classification and U-Net for pixel-level segmentation. This dual-network approach enhances the accuracy of tumor region delineation, which is particularly challenging in some cancer types.

Cell Detection and Classification: Following tumor region identification, a YOLO-based cell detection model computes PD-L1 expression on different cell types for CPS calculation [58]. This model performs triple-task recognition: detection of PD-L1+ tumor cells, PD-L1- tumor cells in tumor regions, and detection of PD-L1+ immune cells in associated non-tumor regions.

Quantitative Continuous Scoring (QCS): A more advanced approach presented by [41] involves quantitative continuous scoring of PD-L1 expression intensity at a granular cell level. This system captures the percentage of tumor cells with medium to strong staining intensity (PD-L1 QCS-PMSTC) and classifies patients with ≥0.575% as biomarker-positive. Unlike traditional binary classification (positive/negative), QCS measures continuous membrane staining intensity values, allowing for more precise patient stratification [41].

Whole-Slide Image Analysis: Most AI systems operate on digitized whole-slide images. The QuPath software comparison study evaluated two protocols: Selected Area Interpretation (AI-SAI) and Whole Slide Imaging (AI-WSI) [104]. AI-SAI demonstrated stronger agreement with manual assessment (κ=0.86) compared to AI-WSI (κ=0.65), suggesting that focused analysis on representative regions may sometimes outperform whole-slide analysis, particularly in bladder cancer cases with high PD-L1-positive tumor cell content [104].

AI PD-L1 Scoring Workflow

Key Technological Approaches

Algorithm Architectures and Innovations

Table 3: AI Algorithm Architectures for PD-L1 Scoring

Algorithm Component	Architecture	Function	Application Example
Cell Detection	YOLOv5/YOLO-based	Localizes and classifies individual cells	PD-L1 positive/negative tumor cell detection [58] [105]
Tissue Segmentation	U-Net	Pixel-level segmentation of tumor regions	Delineating tumor vs. non-tumor areas [58]
Patch Classification	MobileNet-v2	Patch-level classification of tumor regions	Identifying tumor-containing patches [58]
Foundation Model	Vision Transformer (ViT)	Generates image embeddings for classification	FGFR alteration prediction in bladder cancer [107]
Intensity Quantification	Custom computer vision	Measures continuous membrane staining intensity	PD-L1 QCS-PMSTC scoring [41]

Emerging Capabilities

Advanced AI systems now demonstrate capabilities beyond simple replication of pathologist scoring:

Spatial Biomarkers: Researchers from Stanford University developed an AI spatial biomarker that analyzes interactions between tumor cells, fibroblasts, T-cells, and neutrophils [107]. This five-feature model achieved a hazard ratio of 5.46 for progression-free survival in NSCLC patients treated with immune checkpoint inhibitors, significantly outperforming PD-L1 tumor proportion scoring alone (HR=1.67) [107].

Molecular Prediction: Foundation models trained on large WSI datasets can predict molecular alterations directly from H&E-stained slides. Johnson & Johnson's MIA:BLC-FGFR algorithm predicts FGFR alterations in non-muscle invasive bladder cancer with 80-86% AUC, demonstrating strong concordance with traditional molecular testing [107].

Multimodal Integration: Researchers from UCSF and Artera validated a multimodal AI biomarker that combines H&E images with clinical variables (age, Gleason grade, PSA) to predict prostate cancer outcomes after radical prostatectomy [107]. This integration of image-based AI with clinical data improves prognostic tools for personalized treatment strategies.

Scoring Approach Comparison

Essential Research Reagents and Materials

Table 4: Key Research Reagent Solutions for PD-L1 Assessment Studies

Reagent/Platform	Type	Primary Function	Application Notes
PD-L1 IHC Assays (22C3, 28-8, SP263, SP142)	Immunohistochemistry Antibodies	Detection of PD-L1 protein expression	Different clones have specific scoring guidelines and regulatory approvals [103] [58]
Whole Slide Scanners (PANORAMIC1000, Ventana DP200)	Digital Pathology Hardware	Digitization of pathology slides for AI analysis	Resolution (0.25-0.475 μm/pixel) critical for cell-level analysis [9] [58]
QuPath Software	Open-source Digital Pathology Platform	AI-powered cell detection and classification	Offers both Selected Area and Whole Slide analysis protocols [104]
Visiopharm PD-L1 Lung Cancer TME	Commercial AI Application	Automated TPS scoring for NSCLC	Demonstrated substantial agreement with pathologists (κ=0.672) [103]
uPath Software (Roche)	Commercial Digital Pathology Platform	PD-L1 image analysis for SP263 clones	IVDD-certified for TPS ≥50% classification [103] [9]
BenchMark ULTRA Platform	Automated Staining System	Standardized IHC staining procedure	Ensures consistent staining quality for both manual and digital assessment [9] [104]

The comparative analysis of pathologist versus AI algorithm scoring for PD-L1 assessment reveals a complex landscape where both approaches offer complementary strengths. Pathologists demonstrate higher consistency, particularly at critical TPS cutoffs ≥50%, with almost perfect interobserver agreement (κ=0.873) [103]. However, AI algorithms show promising capabilities in quantitative continuous scoring, spatial biomarker analysis, and reducing interobserver variability, particularly in challenging low-expression cases [41] [106].

The emerging paradigm appears to be a collaborative approach where AI algorithms handle quantitative cell detection and classification tasks while pathologists provide contextual interpretation and oversight. This synergistic model leverages the computational power and consistency of AI with the clinical expertise and integrative reasoning of human pathologists. As AI technologies continue to evolve—with advancements in foundation models, multimodal integration, and spatial analysis—they hold significant potential to enhance the precision and predictive power of PD-L1 assessment in both clinical practice and oncology drug development.

For researchers and drug development professionals, the selection between pathologist-based and AI-based scoring approaches should consider factors including the specific cancer type, required throughput, available computational resources, and regulatory requirements. Hybrid models that combine the strengths of both approaches likely represent the future of precision immuno-oncology.

Concordance Studies at Clinically Relevant Cut-offs (1% and 50% for TPS)

Programmed Death-Ligand 1 (PD-L1) immunohistochemistry (IHC) has emerged as a critical predictive biomarker for immune checkpoint inhibitor (ICI) therapy in non-small cell lung cancer (NSCLC) and other malignancies [108] [109]. The Tumor Proportion Score (TPS), which represents the percentage of viable tumor cells exhibiting partial or complete membranous PD-L1 staining, serves as a fundamental scoring metric for determining patient eligibility for immunotherapy [9]. Clinically relevant TPS cut-offs (≥1% and ≥50%) have been established through pivotal clinical trials and are integrated into treatment guidelines worldwide, directly influencing therapeutic decision-making [9] [71].

The development of multiple PD-L1 IHC assays, each with distinct antibody clones and staining platforms, has created significant challenges for pathology laboratories [65] [71]. With limited resources and tissue availability, laboratories face practical difficulties in implementing all commercially available assays, fueling interest in their potential interchangeability [65] [71]. This comprehensive review synthesizes evidence from key concordance studies to evaluate the analytical comparability of various PD-L1 assays at these critical clinical thresholds, providing researchers and clinicians with evidence-based guidance for assay selection and implementation.

Comparative Performance of PD-L1 Assays

Interassay Concordance

Multiple studies have systematically evaluated the analytical concordance between different PD-L1 IHC assays, with consistent findings across various study designs and geographic regions. The evidence demonstrates that while several assays show high agreement, notable exceptions exist that impact their potential interchangeability.

Table 1: Interassay Concordance at Clinically Relevant TPS Cut-offs

Assay Comparison	Sample Size	% Agreement at ≥1% TPS	% Agreement at ≥50% TPS	Statistical Measure	Study
22C3 vs 28-8	144	82.2%	91.6%	Overall Agreement	[108]
22C3 vs SP263	473	>91% (for first-line treatment criteria)	>91% (for first-line treatment criteria)	Positive/Negative Agreement	[110]
22C3 vs SP142	127 vs 132	Lower than 22C3/28-8 comparison	Lower than 22C3/28-8 comparison	Cohen's Kappa	[108]
CAL10 vs SP263	136	≥94.0% (lower bound of 95% CI)	≥86.2% (lower bound of 95% CI)	Overall Percent Agreement	[111]
22C3 vs 28-8 vs SP263 (Blueprint Phase 2)	81	Highly comparable	Highly comparable	Intraclass Correlation	[109]

The Blueprint (BP) Phase 2 study, a pivotal academic and professional society collaboration, provided compelling evidence regarding assay comparability. This comprehensive analysis of 81 real-world lung cancer specimens evaluated five trial-validated PD-L1 assays (22C3, 28-8, SP142, SP263, and 73-10) and demonstrated that 22C3, 28-8, and SP263 assays showed highly comparable staining characteristics for tumor cell PD-L1 expression [109]. In contrast, the SP142 assay exhibited consistently lower sensitivity for detecting PD-L1 expression on tumor cells, while the 73-10 assay demonstrated higher sensitivity compared to other assays [109].

A systematic review published in 2020, which analyzed 27 qualified studies, corroborated these findings, noting that "decrease in concordance... is seen with use of cut-offs, which hampers interchangeability of PD-L1 immunohistochemistry assays" [71]. This observation is clinically relevant as it highlights the challenges in applying binary classifications (positive/negative) to continuous biological variables, particularly near the critical threshold values.

Predictive Value for Therapeutic Response

Beyond analytical concordance, the relationship between assay performance and predictive value for treatment response is paramount. A 2024 prospective study directly compared three different PD-L1 assays (22C3, 28-8, and SP142) for predicting response to combined chemoimmunotherapy in 70 patients with advanced NSCLC [7]. This investigation revealed that PD-L1 expression determined using the 22C3 assay showed stronger correlation with therapeutic response than either the 28-8 or SP142 assays [7]. Specifically, patients with TPS ≥50% as determined by the 22C3 assay had significantly longer progression-free survival compared to those with TPS <50%, while the other assays did not reveal remarkable differences in objective response rate or progression-free survival [7].

Methodological Approaches in Key Studies

Experimental Protocols

The concordance studies employed rigorous methodological approaches to ensure valid and reproducible comparisons:

Sample Preparation and Staining Protocols: Most studies utilized formalin-fixed, paraffin-embedded (FFPE) tissue sections from NSCLC specimens, with consecutive sections to minimize variability due to tumor heterogeneity [110] [109]. The staining procedures followed manufacturers' recommended protocols for each assay: 22C3 and 28-8 assays were typically performed on Dako Autostainer Link 48 platforms, while SP142 and SP263 assays used Ventana BenchMark ULTRA staining systems [108] [110] [109]. The 73-10 assay protocol was developed by Dako/Agilent for avelumab clinical trials [109].

Scoring Methodologies: PD-L1 expression was assessed by experienced pathologists, often with specific training in PD-L1 interpretation [110] [109]. The Tumor Proportion Score was recorded as either continuous variables (0-100%) or categorized using clinically relevant cut-offs (<1%, 1-49%, ≥50%) [108] [109]. Some studies incorporated digital pathology platforms for scoring validation [9] [109], and the Blueprint Phase 2 study demonstrated high concordance between glass slide and digital image scoring (Pearson correlation >0.96) [109].

Statistical Analysis: Concordance was evaluated using various statistical measures including overall percent agreement, Cohen's kappa coefficient (κ), intraclass correlation coefficient (ICC), and Fleiss' kappa for multiple raters [108] [9] [109]. The kappa statistic interpretation typically followed established guidelines: <0.20 poor, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 good, and 0.81-1.00 excellent agreement [108].

Figure 1: Experimental Workflow for PD-L1 Assay Concordance Studies

The Scientist's Toolkit: Key Research Reagents and Platforms

Table 2: Essential Materials for PD-L1 Concordance Research

Category	Specific Reagents/Platforms	Research Function	Key Characteristics
Commercial Assays	PD-L1 IHC 22C3 pharmDx (Dako/Agilent)	Companion diagnostic for pembrolizumab	TPS scoring; Dako Link 48 platform
	PD-L1 IHC 28-8 pharmDx (Dako/Agilent)	Complementary diagnostic for nivolumab	TPS scoring; Dako Link 48 platform
	VENTANA PD-L1 (SP142) Assay (Ventana/Roche)	Complementary diagnostic for atezolizumab	TC/IC scoring; lower tumor cell sensitivity
	VENTANA PD-L1 (SP263) Assay (Ventana/Roche)	Companion diagnostic for durvalumab	Comparable to 22C3/28-8; Ventana platform
	PD-L1 IHC 73-10 Assay (Dako/Agilent)	Developed for avelumab trials	Higher sensitivity; Dako platform
Laboratory-Developed Tests	Cross-platform LDTs (e.g., 22C3 on Ventana)	Increase testing accessibility	Requires proper validation [65]
Staining Platforms	Dako Autostainer Link 48	Platform for 22C3, 28-8, 73-10	Closed system with optimized protocols
	VENTANA BenchMark ULTRA	Platform for SP142, SP263	Automated staining with integrated detection
	BOND-III System (Leica)	Platform for novel assays (e.g., CAL10)	Emerging platform for PD-L1 testing [111]
Digital Pathology	Whole Slide Imaging Scanners	Digital scoring and archiving	Enables remote assessment and AI applications [9]
	AI-Powered Analysis Software	Automated TPS scoring	Shows promise but requires refinement [9]

Additional Considerations in PD-L1 Testing

Interobserver and Interlaboratory Variability

Beyond interassay differences, interobserver variability among pathologists represents another dimension of complexity in PD-L1 testing. Recent studies evaluating both manual and digital scoring demonstrate that interobserver agreement is generally higher at the ≥50% TPS cut-off (almost perfect agreement, Fleiss' kappa 0.873) compared to the ≥1% TPS cut-off (moderate agreement, Fleiss' kappa 0.558) [9]. This finding has significant clinical implications, as discordance at the 1% threshold may potentially impact patient eligibility for certain immunotherapies.

Intraobserver consistency typically remains high (Cohen's kappa 0.726-1.0) [9], suggesting that individual pathologists generally maintain consistent scoring approaches. The Blueprint Phase 2 study further confirmed very strong reliability among pathologists in tumor cell PD-L1 scoring across all assays (overall ICC = 0.86-0.93), though reliability for immune cell scoring was considerably lower (overall ICC = 0.18-0.19) [109].

Emerging Technologies and Assays

The landscape of PD-L1 testing continues to evolve with the development of novel assays and technological approaches. The recently developed PD-L1 CAL10 assay (Leica Biosystems) demonstrated strong concordance with the SP263 assay in a feasibility study, with overall percent agreement lower bounds of 94.0% at ≥1% TPS and 86.2% at ≥50% TPS [111]. This performance suggests that properly validated novel assays can achieve comparable results to established platforms.

Artificial intelligence applications in PD-L1 scoring show promise but currently exhibit limitations. Comparative studies between pathologists and AI algorithms reveal that while some AI tools demonstrate substantial agreement with median pathologist scores (Fleiss' kappa 0.672), their performance remains less consistent than expert human evaluation, particularly at critical clinical decision-making thresholds [9].

Figure 2: PD-1/PD-L1 Signaling Pathway and Clinical Measurement

The comprehensive analysis of PD-L1 IHC assay concordance studies reveals a consistent pattern: the 22C3, 28-8, and SP263 assays demonstrate high analytical comparability for tumor cell scoring at both the 1% and 50% TPS thresholds, suggesting potential interchangeability in clinical practice [108] [109] [71]. In contrast, the SP142 assay shows systematically lower sensitivity for tumor cell staining, while the 73-10 assay exhibits higher sensitivity compared to other assays [109].

These findings have significant implications for diagnostic laboratories and clinical practice. The demonstrated concordance between 22C3, 28-8, and SP263 assays provides flexibility for laboratories in selecting testing platforms based on available infrastructure and resource considerations. However, the predictive superiority of the 22C3 assay for response to chemoimmunotherapy in one prospective study suggests that not all analytically comparable assays may be clinically equivalent [7].

Future directions in PD-L1 testing should focus on standardizing scoring criteria, reducing interobserver variability through enhanced training and potentially artificial intelligence integration, and validating laboratory-developed tests that increase accessibility while maintaining analytical performance. As immunotherapy continues to evolve across cancer types, ensuring accurate, reproducible, and accessible PD-L1 testing remains paramount for optimizing patient selection and treatment outcomes.

Conclusion

The comparative performance of PD-L1 IHC assays reveals a complex landscape where no single assay universally outperforms others, but significant differences in analytical performance and diagnostic accuracy exist. Key takeaways include the critical importance of rigorous validation per CAP guidelines, the moderate-to-poor interchangeability of many assays—particularly at clinically crucial cut-offs—and the persistent challenge of inter-observer variability. Future directions must focus on harmonizing scoring systems, standardizing pre-analytical conditions, and integrating artificial intelligence to enhance reproducibility. The emergence of AI algorithms shows promise but currently lacks the consistency of expert pathologists, highlighting a need for further refinement. For researchers and drug developers, these findings underscore the necessity of fit-for-purpose assay selection and validation to ensure reliable patient stratification in clinical trials and ultimately, optimize immunotherapy outcomes. The evolution of PD-L1 testing will continue to be pivotal in advancing precision immuno-oncology.

References

• 1. New dimensions of PD-1/PD-L1 inhibitor combination ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12420611/]
• 2. Overcoming resistance to anti-PD-L1 immunotherapy [https://molecular-cancer.biomedcentral.com/articles/10.1186/s12943-025-02400-z]
• 3. The mechanism and research progress of PD-1/PD-L1 on ... [https://pubmed.ncbi.nlm.nih.gov/41158269/]
• 4. Immune evasion in cancer: mechanisms and cutting-edge ... [https://www.nature.com/articles/s41392-025-02280-1]
• 5. Advances in PD-1/PD-L1 pathway inhibitors in the ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1643421/full]
• 6. “Interchangeability” of PD-L1 immunohistochemistry assays [https://www.nature.com/articles/s41379-019-0327-4]
• 7. Comparing Three Different Anti–Programmed Death ... [https://www.sciencedirect.com/science/article/pii/S2666364324000146]
• 8. 2025 MIPS Measure #506: Positive PD-L1 Biomarker ... [https://mdinteractive.com/mips_quality_measure/2025-mips-quality-measure-506]
• 9. Comparative performance of PD‐L1 scoring by ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12129605/]
• 10. Cancer Biomarkers: Reflection on Recent Progress, Emerging ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12468363/]
• 11. Predictive models for immune checkpoint inhibitor ... [https://www.sciencedirect.com/science/article/pii/S1040842825003683]
• 12. Immune Biomarkers for Checkpoint Blockade in Solid Tumors [https://pmc.ncbi.nlm.nih.gov/articles/PMC12384685/]
• 13. Companion diagnostics and predictive biomarkers for PD-1 ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2024.1454720/full]
• 14. Recent advances in biomarkers for predicting the efficacy ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1554871/full]
• 15. Evaluating the significance of combining PD-L1 and TILs ... [https://www.nature.com/articles/s44276-025-00174-x]
• 16. PD-L1 Biomarker Testing Market Trends & Forecast 2025– ... [https://www.futuremarketinsights.com/reports/pd-l1-biomarker-testing-market]
• 17. A comparability study of immunohistochemical assays for ... [https://www.sciencedirect.com/science/article/pii/S0893395222009887]
• 18. Comparison of different predictive biomarker testing assays ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2023.1265202/full]
• 19. Exosomal PD-L1 detection in cancer predictive biomarker ... [https://pubmed.ncbi.nlm.nih.gov/40677702/]
• 20. Current status and progress of PD-L1 detection [https://pmc.ncbi.nlm.nih.gov/articles/PMC11258158/]
• 21. PD-1 and PD-L1 Inhibitors Competitive Landscape 2025 ... [https://finance.yahoo.com/news/pd-1-pd-l1-inhibitors-152000823.html]
• 22. Companion diagnostic assays for PD-1/PD-L1 checkpoint ... [https://pubmed.ncbi.nlm.nih.gov/26559787/]
• 23. Companion and Complementary Diagnostics-Focus on PD ... [https://pubmed.ncbi.nlm.nih.gov/29084031/]
• 24. PD-L1 Companion Diagnostics and Scoring Systems [https://genomoncology.com/blog/pd-l1-companion-diagnostics-and-scoring-systems/]
• 25. Evaluation of the diagnostic concordance of FDA-approved ... [https://www.nature.com/articles/s41598-025-05697-4]
• 26. Comparison of different predictive biomarker testing assays ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10562577/]
• 27. Validation of E1L3N antibody for PD-L1 detection and ... [https://www.nature.com/articles/s43856-022-00206-4]
• 28. PD-L1, TMB, MSI, and Other Predictors of Response to ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7867133/]
• 29. Tumor Mutational Burden (TMB) as a Predictive Biomarker ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7710563/]
• 30. Biology and evolving management of resectable dMMR ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12592604/]
• 31. Comparison of standard mismatch repair deficiency and ... [https://translational-medicine.biomedcentral.com/articles/10.1186/s12967-024-04960-y]
• 32. A scoping review and meta-analysis on the prevalence of ... [https://www.nature.com/articles/s41598-022-23319-1]
• 33. Clinicopathological characteristics and the relationship of PD ... [https://bmccancer.biomedcentral.com/articles/10.1186/s12885-025-13938-y]
• 34. Uncovering PD-L1 among immune cell predictive ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12182785/]
• 35. Chinese Experts Consensus on Immune Checkpoint ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8105610/]
• 36. Sources of inter-individual variability leading to significant ... [https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2022.1056365/full]
• 37. Why PD-L1 expression varies between studies of lung cancer [https://www.nature.com/articles/s41598-024-80301-9]
• 38. 肿瘤细胞PD-L1表达的阈值究竟在哪 [https://tlcr.amegroups.org/article/view/28754/48581]
• 39. Analytical validation and initial clinical testing of quantitative ... [https://biomarkerres.biomedcentral.com/articles/10.1186/s40364-022-00370-8]
• 40. Intrapatient variation in PD-L1 expression and tumor ... [https://www.sciencedirect.com/science/article/pii/S0923753424007452]
• 41. Enhanced patient selection with quantitative continuous ... [https://www.nature.com/articles/s41698-025-01107-0]
• 42. PD-L1 Testing, Treatment Patterns, and Clinical Outcomes ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11998521/]
• 43. Recent Advancement of PD-L1 Detection Technologies and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10088895/]
• 44. Role of PD-L1 SP263 Assay in Cancer Immunotherapy [https://www.scientificarchives.com/article/role-of-pd-l1-sp263-assay-in-cancer-immunotherapy-a-comprehensive-review]
• 45. Comparison of SP263 and 22C3 immunohistochemistry PD ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10619123/]
• 46. PD-L1 testing using the clone 22C3 pharmDx kit for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5919255/]
• 47. PD-L1 Immunohistochemistry Assay Comparison in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8634452/]
• 48. Standardization of PD-L1 immunohistochemistry [https://www.nature.com/articles/s41379-021-00917-4]
• 49. Quantitative Assessment of PD-L1 as an Analyte in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6920558/]
• 50. Comparative evaluation of automated systems in ... [https://www.sciencedirect.com/science/article/pii/S0009898198001466]
• 51. BenchMark ULTRA IHC/ISH System [https://diagnostics.roche.com/us/en/products/instruments/benchmark-ultra-ins-2099.html]
• 52. BOND-III Fully Automated IHC and ISH Staining System [https://www.leicabiosystems.com/us/ihc-ish/ihc-ish-instruments/bond-iii/]
• 53. Two Instrument Comparison of Reagents From a US FDA ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9444283/]
• 54. Comparison of MIB‐1‐Specific Membrane Staining in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11584304/]
• 55. Analytical Comparison of Commonly Used Laboratory ... [https://www.sciencedirect.com/science/article/pii/S0893395225001097]
• 56. Development of a Novel IHC Assay for PD-L1 Detection ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12562258/]
• 57. Comparison of PD-L1 expression in head and neck ... [https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1666078/full]
• 58. Automated interpretation of PD-L1 CPS based on multi-AI ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12364678/]
• 59. Prediction of a Panel of Programmed Cell Death Protein-1 (PD ... [https://cancer.jmir.org/2025/1/e67379/]
• 60. Decoding the molecular landscape: HER2 and PD-L1 in ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1567308/full]
• 61. PD-L1 expression assessment in Angiosarcoma improves ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12166781/]
• 62. FDA Takes Action Aimed at Helping to Ensure the Safety ... [https://www.fda.gov/news-events/press-announcements/fda-takes-action-aimed-helping-ensure-safety-and-effectiveness-laboratory-developed-tests]
• 63. Public Health Implications of Legal Challenges to the ... [https://pubmed.ncbi.nlm.nih.gov/40540283/]
• 64. Federal Judge Vacates FDA's Final Rule Reclassifying ... [https://www.hipaajournal.com/federal-judge-vacates-fda-final-rule-laboratory-developed-tests/]
• 65. Comparability of laboratory-developed and commercial PD ... [https://www.sciencedirect.com/science/article/abs/pii/S1092913420301337]
• 66. Demystifying the Development and Implementation of ... [https://www.thermofisher.com/blog/clinical-conversations/demystifying-the-development-and-implementation-of-molecular-tests-in-a-clinical-laboratory/]
• 67. The FDA's Proposed Ruling on Lab Tests Could Have ... [https://medicine.yale.edu/news-article/the-fdas-proposed-ruling-on-lab-tests-could-have-unintended-consequences/]
• 68. A Product Development Plan For My Laboratory ... [https://namsa.com/resources/blog/a-product-development-plan-for-my-laboratory-developed-test/]
• 69. Quantitative comparison of PD-L1 IHC assays against NIST ... [https://pubmed.ncbi.nlm.nih.gov/34389791/]
• 70. Demonstrating the interference of tissue processing in ... [https://www.sciencedirect.com/science/article/pii/S0344033823003059]
• 71. Comparability of PD‐L1 immunohistochemistry assays for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7318295/]
• 72. Quantitative comparison of PD-L1 IHC assays against NIST ... [https://www.sciencedirect.com/science/article/pii/S089339522200309X]
• 73. Understanding Preanalytical Variables and their Effects on ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6004232/]
• 74. Decline of PD-L1 Immunoreactivity with Storage Duration ... [https://www.mdpi.com/2072-6694/17/19/3103]
• 75. Impact of delayed and prolonged fixation on the evaluation ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6647403/]
• 76. of different predictive biomarker testing Comparison for... assays [https://pubmed.ncbi.nlm.nih.gov/37822932/]
• 77. Predicability of PD-L1 expression in cancer cells based ... [https://www.sciencedirect.com/science/article/pii/S2153353925001105]
• 78. PD-L1 antibodies: variation in therapeutic and detection uses [https://www.abcam.com/en-us/stories/articles/pd-l1-antibodies]
• 79. Inter- and Intra-Observer Agreement of PD-L1 SP142 ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10000421/]
• 80. Interassay and interobserver comparability study of four ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8602040/]
• 81. Spatial heterogeneity of PD-L1 expression influence its ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12205590/]
• 82. Heterogeneity-optimized method for predicting immune ... [https://www.nature.com/articles/s41598-025-17718-3]
• 83. Exosomal PD-L1 detection in cancer predictive biomarker ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1603855/full]
• 84. Cost-effectiveness of PD-L1 testing to guide ... [https://pubmed.ncbi.nlm.nih.gov/41235512/]
• 85. Spatiotemporal analyses of the pan-cancer single-cell ... [https://www.nature.com/articles/s43018-025-01039-5]
• 86. Clinical evaluation of an automated pan-organ combined ... [https://www.sciencedirect.com/science/article/pii/S2949820125000700]
• 87. A novel deep learning framework for automatic scoring of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12499557/]
• 88. BenchMark Special Stains system [https://diagnostics.roche.com/us/en/products/instruments/benchmark-special-stains-ins-3198.html]
• 89. Optimisation of a Standardised Automated Bleaching and ... [https://pubmed.ncbi.nlm.nih.gov/41168914/]
• 90. Color Normalization in Breast Cancer Immunohistochemistry ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12468545/]
• 91. Chromogenic IHC and ISH Resources [https://diagnostics.roche.com/us/en/products/product-category/immunohistochemistry--ihc-/chromogenic-resources.html]
• 92. Intro to Immunohistochemistry (IHC) Staining: Steps & Best ... [https://www.leicabiosystems.com/us/knowledge-pathway/immunohistochemistry-an-overview-steps-to-better-ihc-staining/]
• 93. Counterstains in Immunohistochemistry: Principles, ... [https://www.bosterbio.com/blog/post/counterstains-in-immunohistochemistry-principles-options-and-troubleshooting?srsltid=AfmBOoqn8doh-9g6Aual5XAV5qjxPBYOhbwSyEeVByCMM7DAAQAbj--s]
• 94. æ–°åž‹PD-L1检测抗体3E2在肺腺癌中的临床评估 [https://www.ebiotrade.com/newsf/2025-10/20251018004342852.htm]
• 95. 既往治疗过的非小细胞肺癌的免疫治疗 [https://jtd.amegroups.org/article/view/18966/46393]
• 96. 免疫疗法：新型抗PD-L1 抗体与各种组合方案 [https://memoinoncology.com/%E5%85%8D%E7%96%AB%E7%96%97%E6%B3%95%EF%BC%9A%E6%96%B0%E5%9E%8B%E6%8A%97-pd-l1-%E6%8A%97%E4%BD%93%E4%B8%8E%E5%90%84%E7%A7%8D%E7%BB%84%E5%90%88%E6%96%B9%E6%A1%88/]
• 97. 免疫组织化学市场规模、趋势、竞争格局2025 - 2030 [https://www.mordorintelligence.com/zh-CN/industry-reports/immunohistochemistry-market]
• 98. 探索多重免疫组织化学相对于传统IHC 的优势 [https://www.celnovte.com/zh-CN/solution/exploring-the-benefits-of-multiplex-immunohistochemistry-over-conventional-ihc/]
• 99. The Ring Study: an international comparison of PD - L diagnostic... 1 [https://pubmed.ncbi.nlm.nih.gov/36319485/]
• 100. High-Sensitivity PD-L1 Staining Using Clone 73−10 ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12092928/]
• 101. Different PD-L1 Assays Reveal Distinct Immunobiology and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11962399/]
• 102. PD-L1 immunohistochemical expression in bladder ... [https://www.sciencedirect.com/science/article/abs/pii/S1092913424000042]
• 103. Comparative performance of PD-L1 scoring by pathologists ... [https://pubmed.ncbi.nlm.nih.gov/39961605/]
• 104. Artificial Intelligence (AI) for Programmed Death Ligand-1 ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12194390/]
• 105. Comparing deep learning and pathologist quantification of ... [https://www.nature.com/articles/s41598-024-57067-1]
• 106. Clinical Validation of Artificial Intelligence–Powered PD-L1 ... [https://www.lunit.io/en/publication/clinical-validation-of-artificial-intelligence-powered-pd-l1-tumor-proportion-score-interpretation-for-immune-checkpoint-inhibitor-response-prediction-in-non-small-cell-lung-cancer/]
• 107. Digital Pathology and AI Highlights from ASCO 2025 [https://proscia.com/digital-pathology-and-ai-highlights-from-asco-2025/]
• 108. PD-L1 Immunohistochemistry Comparability and Their ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7655253/]
• 109. PD-L1 Immunohistochemistry Comparability Study in Real ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8386299/]
• 110. Agreement between PDL1 immunohistochemistry assays ... [https://www.nature.com/articles/s41598-020-60950-2]
• 111. Development of a Novel IHC Assay for PD-L1 Detection in ... [https://www.mdpi.com/2227-9059/13/10/2359]