Stage-aware programmed death ligand 1 assessment in gastric cancer: Implications for perioperative immunotherapy

Abstract

Programmed death ligand 1 expression in gastric cancer is highly stage-dependent, influencing both prognosis and perioperative immune checkpoint inhibitor response. Conventional combined positive score assessment is limited by spatial heterogeneity, cellular source ambiguity, and dynamic changes under treatment. This opinion review synthesizes current advances in stage-aware programmed death ligand 1 evaluation, including spatial profiling, multi-omics integration, microbiota analysis, and longitudinal monitoring. We emphasize integrating tumor stage, immune microenvironment, and molecular features to enable context-dependent, precision immunotherapy. By highlighting ongoing controversies and emerging strategies, this review provides actionable insights for biomarker-driven trial design and clinical decision-making, and outlines future directions to optimize perioperative management of gastric cancer.

Key Words: Programmed death ligand 1; Gastric cancer; Stage-aware assessment; Immune checkpoint inhibitors; Tumor microenvironment; Multi-omics integration; Spatial profiling; Perioperative immunotherapy; Context-dependent biomarkers; Precision oncology

Core Tip: This opinion review emphasizes the stage- and context-dependent interpretation of programmed death ligand 1 (PD-L1) in gastric cancer. We summarize current advances in PD-L1 assessment, including combined positive score, spatial profiling, and multi-omics integration, and discuss how tumor stage, immune microenvironment, and T-cell functionality influence its prognostic and predictive value. The review provides a framework for integrating PD-L1 into perioperative immunotherapy, highlighting adaptive, biomarker-guided strategies to optimize patient selection and therapeutic outcomes. By bridging molecular insights with personalized treatment, this work offers practical guidance for both clinicians and researchers in precision gastric cancer management.

INTRODUCTION

Gastric cancer (GC) remains a leading cause of cancer-related mortality worldwide, with particularly high incidence in East Asia[1-4]. Despite advances in surgical techniques and perioperative therapies, including chemotherapy and radiotherapy, recurrence and metastasis continue to pose major challenges for patients with resectable disease[1,5]. In recent years, immune checkpoint inhibitors (ICIs) targeting the programmed death-1/programmed death ligand 1 (PD-L1) axis have emerged as transformative agents, demonstrating clinically meaningful survival benefits in selected patients with advanced or metastatic GC[3,6-8]. While these advances underscore the potential of immunotherapy, translating such benefits to the perioperative setting remains complex, largely due to heterogeneity in tumor biology, immune microenvironment, and treatment response[9-11].

PD-L1 has been widely adopted as a predictive and prognostic biomarker in GC, most commonly quantified using the combined positive score (CPS), which integrates PD-L1 expression on both tumor cells and tumor-infiltrating immune cells[12-16]. While CPS has facilitated patient selection in clinical trials, several studies report inconsistent associations between PD-L1 expression and clinical outcomes, particularly in localized or resectable disease[6,17-19]. These discrepancies likely reflect intrinsic biological variability, including differences in tumor mutational burden (TMB), microsatellite instability (MSI), Epstein-Barr virus (EBV) status, and immune microenvironment composition, as well as technical limitations such as inter-observer variability, platform differences, and sampling bias[7,13,18]. From an author perspective, these factors underscore the need for caution when interpreting CPS and highlight the importance of context-aware evaluation.

Emerging evidence demonstrates that pathologic stage critically modulates both PD-L1 expression and its functional significance. Progression from stage II to stage III is often accompanied by profound changes in the tumor microenvironment, including increased infiltration of immunosuppressive myeloid cells, expansion of M2 macrophages, and heightened T-cell exhaustion, collectively enhancing PD-L1-mediated immune evasion[6,20-23]. These stage-dependent dynamics suggest that PD-L1 should not be interpreted in isolation. Instead, a context-aware assessment that integrates tumor stage, immune composition, and cellular origin is essential for accurate prognostication and therapeutic planning[24-26]. For example, in stage III GC, high PD-L1 expression alongside abundant M2 macrophages may indicate limited ICI responsiveness, whereas similar CPS values in stage II tumors may reflect more effective immune surveillance[6,20].

In this opinion review, we provide a comprehensive overview of current advances, ongoing controversies, and clinical implications of PD-L1 assessment in GC, emphasizing stage-dependent interpretation. We discuss limitations of conventional biomarker strategies, highlight emerging technologies, including spatial profiling, multi-omics integration, and microbiota analysis[26-30], and outline future directions to optimize perioperative immunotherapy. By framing PD-L1 evaluation within a stage- and context-aware paradigm, we aim to provide actionable insights to guide both clinical practice and future translational research[10,11].

PD-L1 ASSESSMENT IN GC

PD-L1 has emerged as a central biomarker in GC, reflecting its dual role in modulating tumor immune evasion and predicting response to ICIs[6,13,17]. Accurate quantification of PD-L1 is critical for both prognostic evaluation and therapeutic decision-making, particularly in perioperative immunotherapy trials[8,31]. The CPS, which integrates PD-L1 expression on tumor cells and tumor-infiltrating immune cells, remains the most widely used metric[13,17]. However, despite its broad adoption, CPS presents notable limitations. Multi-center studies report inter-observer variability, differences in staining platforms, and inconsistent scoring thresholds, leading to variable correlations between CPS and clinical outcomes[16,18,19,32]. These methodological challenges underscore the need for standardized, robust PD-L1 assessment strategies that can be reliably applied in perioperative settings.

Recent technological innovations are beginning to address these limitations. Artificial intelligence-assisted image analysis and digital pathology platforms enable automated, reproducible quantification of PD-L1 across heterogeneous tumor regions[17,26]. These approaches reduce human bias, enhance inter-laboratory consistency, and provide more precise evaluation of spatial heterogeneity, which is critical for understanding PD-L1 function in the tumor microenvironment. Moreover, integrating PD-L1 assessment with complementary biomarkers, including TMB, MSI, and EBV status, can improve predictive accuracy of CPS, particularly in early-stage or perioperative contexts[3,18,28]. From an author perspective, such multi-parametric strategies are essential not only for refined patient selection but also for designing biomarker-driven clinical trials that account for tumor and immune complexity.

In resectable or locally advanced GC, PD-L1 demonstrates variable prognostic significance. Some studies indicate that higher PD-L1 expression correlates with enhanced response to ICIs, reflecting an immune-activated tumor microenvironment[31]. Conversely, elevated PD-L1 may associate with more aggressive tumor biology and poorer outcomes in subsets of patients, particularly when accompanied by immunosuppressive myeloid infiltration, M2 macrophages, or T-cell exhaustion[6,20,23]. This variability reinforces the complexity of interpreting PD-L1 and highlights the necessity of integrating stage- and context-dependent analyses. From an author perspective, CPS alone cannot fully capture functional immune dynamics, and its interpretation should be informed by tumor microenvironment characteristics.

Furthermore, perioperative clinical trials increasingly incorporate PD-L1 assessment into biomarker-guided patient stratification, demonstrating its practical relevance for clinical decision-making[18,33]. These studies illustrate that combining PD-L1 evaluation with immune context, molecular subtype, and disease stage improves treatment selection, enhances response prediction, and identifies patients most likely to benefit from ICIs, while minimizing exposure to ineffective therapy[8,34,35]. From an author perspective, future trial designs should explicitly integrate these multi-dimensional parameters to achieve precision perioperative immunotherapy.

Taken together, current advances in PD-L1 assessment highlight a clear trend toward more quantitative, standardized, and integrated evaluation strategies. Combining CPS with artificial intelligence-assisted pathology, complementary molecular biomarkers, and context-aware interpretation provides a foundation for precision patient stratification and rational trial design. Nevertheless, challenges remain in reconciling methodological variability, biological heterogeneity, and stage-dependent differences, underscoring the necessity for ongoing refinement of assessment approaches and prospective multi-center validation[25,26,36]. Embedding these strategies into perioperative practice represents a critical step toward context-aware, stage-specific PD-L1 evaluation that can inform both clinical care and translational research.

STAGE-DEPENDENT SIGNIFICANCE OF PD-L1

Accumulating evidence indicates that the prognostic and predictive value of PD-L1 in GC varies across disease stages, highlighting the need for a stage-aware interpretation. Tumors at different pathologic stages exhibit distinct immune microenvironments, which substantially influence PD-L1 expression, spatial distribution, and functional impact[6,12]. For instance, stage III tumors frequently display higher intratumoral heterogeneity, accelerated clonal evolution, and a more immunosuppressive milieu, including increased infiltration of M2-polarized macrophages, myeloid-derived suppressor cells (MDSCs), and cancer-associated fibroblasts (CAFs), collectively enhancing PD-L1-mediated immune evasion[6,21,37]. In contrast, stage II tumors generally retain a more functional and polyclonal T-cell repertoire, with PD-L1 expression primarily reflecting basal immune regulation rather than adaptive immune resistance[25,26,38].

Figure 1 illustrates stage-dependent variations in PD-L1 expression and immune cell composition, highlighting differences in T-cell functionality, MDSC and M2 macrophage abundance, and CAF activity. Table 1 summarizes key immune microenvironment differences between stage II and stage III GC, emphasizing their implications for PD-L1 functional relevance. From an author perspective, these comparisons underscore that PD-L1 interpretation must be contextualized by tumor stage and microenvironmental features to accurately inform prognosis and treatment planning.

Figure 1 Stage-dependent variations in programmed death ligand 1 expression and immune cell composition. MDSC: Myeloid-derived suppressor cell; PD-L1: Programmed death ligand 1; CAF: Cancer-associated fibroblasts.

Table 1 Stage-specific immune microenvironment and programmed death ligand 1 interpretation in gastric cancer.

Feature1	Stage II	Stage III	Implication for PD-L1 interpretation
CD8+ T-cell function	Functional/high	Exhausted/Low	In stage II, PD-L1 may reflect basal immune regulation; in stage III, PD-L1 indicates immune escape potential
MDSC infiltration	Low	High	Higher immunosuppression in stage III amplifies PD-L1-mediated immune evasion
M2 macrophages	Low	High	Suppressive macrophages in stage III contribute to T-cell inhibition
Cancer-associated fibroblasts	Moderate	High	Dense stroma in stage III limits immune infiltration, enhancing PD-L1 functional impact
PD-L1 expression	Baseline/moderate	Adaptive upregulation/high	PD-L1’s prognostic and predictive relevance is stage-dependent
Spatial heterogeneity	Low-moderate	High	Stage III tumors show more intratumoral variation in PD-L1 and immune cells
T-cell exhaustion markers	Low	High	Correlates with functional impairment of anti-tumor immunity in stage III

¹Comparison of CD8+ T-cell function, myeloid-derived suppressor cell and M2 macrophage infiltration, cancer-associated fibroblast density, programmed death ligand 1 expression levels, spatial heterogeneity, and T-cell exhaustion markers between stage II and III tumors. These features collectively illustrate the stage-dependent prognostic and predictive significance of programmed death ligand 1.

MDSC: Myeloid-derived suppressor cell; PD-L1: Programmed death ligand 1.

Tumor evolution and immune editing

Progression from early to late stages in GC is associated with clonal selection and immune editing, which shape both the level and spatial distribution of PD-L1[6,12]. Stage III tumors often harbor subclonal populations exhibiting adaptive PD-L1 upregulation, driven by selective pressures from cytotoxic T lymphocytes and other immune effectors, as demonstrated in multi-omics analyses of neoadjuvant-treated tumors[27,28,39]. These adaptive changes correlate with enhanced immune evasion and more aggressive tumor biology. Conversely, PD-L1 expression in stage II tumors largely reflects baseline immune modulation with minimal selective pressure, resulting in weaker associations with adverse outcomes. Consequently, identical CPS may carry distinct prognostic meanings depending on tumor stage, highlighting the importance of stage-specific interpretation.

Myeloid remodeling and suppressive immune ecology

The functional relevance of PD-L1 is strongly influenced by tumor microenvironment composition, particularly myeloid populations and stromal elements. MDSCs and M2 macrophages inhibit T-cell cytotoxicity and secrete immunosuppressive cytokines, while CAFs remodel the extracellular matrix and restrict immune infiltration[20,26]. In stage III tumors, the density and suppressive activity of these populations are higher, potentiating PD-L1-mediated immune escape. In contrast, stage II tumors often maintain a less suppressive microenvironment, allowing PD-L1-positive T cells to exert more effective immune surveillance. From an author standpoint, these observations reinforce that PD-L1 cannot be evaluated in isolation; integrating microenvironmental parameters is essential for functional interpretation and trial design.

Functional state of T cells

The phenotypic and functional status of tumor-infiltrating lymphocytes critically shapes PD-L1 interpretation. Stage-dependent variations in T-cell exhaustion, activation, and spatial organization can result in divergent functional consequences, even with comparable CPS measurements[12,20,40]. Stage III tumors frequently harbor a higher proportion of exhausted CD8+ T cells and regulatory T cells, which may blunt anti-tumor immunity and amplify PD-L1’s role as a suppressive checkpoint. In stage II tumors, tumor-infiltrating lymphocytes generally retain higher effector function, and PD-L1 expression may not correlate with immune evasion to the same extent. These findings emphasize that PD-L1 assessment should consider T-cell functionality in conjunction with CPS, rather than relying solely on expression levels.

Stage-dependent implications for biomarker interpretation

Collectively, these findings indicate that PD-L1 is not a static biomarker. Its prognostic and predictive relevance is intimately linked to tumor stage, immune microenvironment composition, and T-cell functionality. Stage-aware interpretation is therefore essential for accurate prognostic assessment and for guiding perioperative immunotherapy. Integrating PD-L1 evaluation with additional biomarkers, such as TMB, MSI, EBV status, and spatial immune context, can improve predictive accuracy and enable personalized treatment strategies[12,26].

From an author perspective, acknowledging stage-dependent differences in tumor-immune interactions not only helps avoid patient misclassification but also provides a foundation for dynamic, context-aware PD-L1 assessment. Such an approach can inform trial design, patient stratification, and adaptive perioperative immunotherapy strategies. We advocate that future studies explicitly integrate tumor stage, immune composition, and T-cell functionality into PD-L1 interpretation to achieve precision perioperative immunotherapy in GC.

CONTROVERSIES IN PD-L1 INTERPRETATION

Despite the widespread adoption of PD-L1 as a biomarker in GC, several critical controversies remain unresolved, complicating both its clinical interpretation and application in perioperative immunotherapy. These controversies span methodological limitations, biological heterogeneity, spatial variability, and dynamic changes under therapy, all of which influence PD-L1’s prognostic and predictive value[13,17,25,41]. From an author perspective, understanding these complexities is essential for advancing stage-aware and context-integrated biomarker evaluation.

Limitations of the CPS

The CPS remains the standard metric for quantifying PD-L1, integrating staining on both tumor cells and tumor-infiltrating immune cells[13,17]. However, CPS has intrinsic limitations: It provides only a global estimate of PD-L1 expression and does not account for spatial distribution or the functional status of immune cells[28,42-44]. Furthermore, inter-observer variability, differences in staining platforms, and discrepancies in scoring thresholds contribute to inconsistent results across studies[15,19,34]. These technical challenges partially explain contradictory associations between CPS and clinical outcomes, particularly in resectable or locally advanced GC. From an author perspective, future biomarker strategies should incorporate both quantitative and spatially resolved metrics to overcome these limitations and improve interpretability.

Cellular source ambiguity

PD-L1 is expressed not only by tumor cells but also by diverse immune populations, including M2-polarized macrophages, MDSCs, and dendritic cells[25,42,43]. The functional significance of PD-L1 may therefore differ depending on its cellular origin. Standard CPS scoring does not differentiate between tumor-derived and immune-derived PD-L1, introducing uncertainty in both prognostic interpretation and therapeutic response prediction. For example, high PD-L1 on immune cells at the invasive margin may indicate active anti-tumor immunity, whereas tumor cell-derived PD-L1 primarily reflects adaptive immune resistance[26,27,45]. From an author perspective, these nuances reinforce the need for cellular-resolution assessment in perioperative trials to accurately guide patient selection and therapeutic decisions.

Spatial heterogeneity

Tumor tissue exhibits pronounced spatial heterogeneity, with PD-L1 expression varying across tumor cores, invasive margins, and tertiary lymphoid structures (TLS)[25,43,46]. High PD-L1 expression in peritumoral TLS may reflect immune activation, whereas diffuse intratumoral expression can indicate immune evasion. Conventional single-site biopsy sampling often fails to capture this heterogeneity, limiting predictive accuracy. We propose that multi-region sampling or digital spatial profiling be incorporated into trial design to better capture these biologically relevant patterns and improve biomarker reliability.

Dynamic changes under therapy

PD-L1 expression is inherently dynamic, influenced by neoadjuvant chemotherapy, immunotherapy, and tumor evolution[14,25,47]. Baseline CPS measurements may not accurately reflect PD-L1 status during the perioperative period, and serial assessments are increasingly recommended to guide adaptive treatment strategies. Ignoring these temporal fluctuations risks patient misclassification and suboptimal therapeutic outcomes. From an author standpoint, longitudinal monitoring should be considered a core component of precision perioperative immunotherapy protocols, enabling dynamic adjustment of treatment strategies.

Integrated, context-aware interpretation

Collectively, these controversies underscore the necessity for an integrated, context-aware PD-L1 assessment framework that incorporates tumor stage, cellular origin, spatial heterogeneity, and temporal dynamics. Addressing these factors is critical for accurate patient stratification, adaptive perioperative immunotherapy, and improved clinical trial design[13,25,42,48]. By combining multi-parametric evaluation with stage-dependent interpretation, clinicians and researchers can better identify patients most likely to benefit from ICIs while minimizing unnecessary exposure and toxicity. From our perspective, integrating spatial, cellular, and longitudinal data ensures precision in biomarker-guided decision-making, moving beyond static CPS toward dynamic, clinically actionable evaluation.

IMPLICATIONS FOR PERIOPERATIVE IMMUNOTHERAPY

The stage-dependent variability and controversies surrounding PD-L1 expression have important implications for the design and implementation of perioperative immunotherapy trials in GC. Recognizing that PD-L1’s prognostic and predictive relevance differs between stage II and stage III tumors, integrating stage-aware biomarker assessment into trial design is essential for accurate patient stratification, improved trial efficiency, and optimized clinical outcomes[8,31]. From an author perspective, failure to account for stage-specific immune contexts may obscure treatment effects and limit the interpretability of trial results. Figure 2 illustrates a framework for integrating PD-L1 assessment into perioperative trial design, encompassing stage-stratified enrollment, preplanned subgroup analyses, biomarker-enriched cohorts, and dynamic biomarker monitoring. These strategies collectively optimize patient selection, enhance therapeutic efficacy, and inform adaptive treatment decisions.

Figure 2 Framework for integrating programmed death ligand 1 assessment into perioperative trial design. MDSC: Myeloid-derived suppressor cell; PD-L1: Programmed death ligand 1; CPS: Combined positive score; EBV: Epstein-Barr virus; MSI: Microsatellite instability; CIN: Chromosomal instability; GS: Genomic stability; TMB: Tumor mutational burden; pCR: Pathological complete response; EFS: Event-free survival; OS: Overall survival; Tx: Treatment.

Stage-stratified enrollment

Tumors at different stages exhibit distinct immune microenvironments that influence baseline PD-L1 expression and responsiveness to ICIs. Stage-stratified enrollment should therefore be implemented in perioperative trials, ensuring separate analysis of stage II and stage III cohorts[8,31,49]. This approach increases the ability to detect stage-specific therapeutic effects and reduces confounding from heterogeneous immune contexts. For instance, stage III tumors with elevated immunosuppressive myeloid infiltration may respond differently than stage II tumors with more functional T-cell repertoires, justifying tailored inclusion criteria. From an author perspective, explicitly integrating stage-based stratification is essential to maximize interpretability of biomarker-driven outcomes.

Preplanned subgroup analyses

To account for intra-stage heterogeneity, trials should incorporate preplanned subgroup analyses based on CPS, PD-L1 cellular origin, and molecular subtypes, such as EBV-positive, MSI-high, or chromosomal instability tumors[14,15,26,50]. These analyses facilitate subpopulation-specific interpretation of therapeutic efficacy and identify patients most likely to benefit from perioperative ICIs. From an author perspective, combining PD-L1 assessment with spatial immune context or TMB can further enhance predictive precision and support rational trial design.

Biomarker-enriched cohorts

Beyond stage stratification, trials may establish biomarker-enriched cohorts, selecting patients based on PD-L1 expression patterns or immune-activated tumor microenvironments, such as high CD8+ T-cell infiltration combined with low MDSC abundance[20,31,51]. This strategy increases the likelihood of demonstrating therapeutic efficacy while minimizing exposure of non-responders to ICIs, thereby enhancing both safety and signal detection. Biomarker enrichment also supports adaptive trial designs, allowing investigators to refine inclusion criteria as emerging immunologic or molecular data become available.

Dynamic biomarker reassessment

Given the dynamic nature of PD-L1 expression, serial evaluation before and after neoadjuvant therapy is recommended[14,26,52]. Longitudinal monitoring provides critical insights into therapy-induced modulation of PD-L1, emerging resistance mechanisms, and residual disease biology. Incorporating dynamic assessment enables adaptive treatment strategies, guides post-surgical adjuvant therapy, and optimizes the timing and selection of perioperative ICIs. From an author perspective, longitudinal evaluation is a key step toward context-aware, stage-specific biomarker integration in clinical practice.

Summary and integration

In summary, integrating stage-aware PD-L1 assessment into perioperative trial design, through stratified enrollment, preplanned subgroup analyses, biomarker-enriched cohorts, and dynamic monitoring, substantially enhances the clinical utility of ICIs in GC. These strategies underscore the importance of context-dependent biomarker application rather than reliance on static baseline CPS measurements[8,15,31,53]. By incorporating tumor stage, immune microenvironment, and molecular features into trial design, clinicians and researchers can optimize patient selection, improve response prediction, and enhance therapeutic efficacy, ultimately laying the groundwork for precision perioperative immunotherapy. From an author perspective, adopting these integrated strategies is critical for translating PD-L1 biology into actionable clinical decision-making.

EMERGING DIRECTIONS

Recent technological innovations and an improved understanding of tumor-immune interactions have highlighted several emerging directions that may enhance the precision and clinical utility of PD-L1 as a biomarker in GC. These approaches focus on capturing spatial, molecular, microbial, and dynamic aspects of the tumor microenvironment, which are increasingly recognized as critical determinants of immune responsiveness and therapeutic efficacy[6,26,54]. From an author perspective, integrating these modalities provides actionable insights that move PD-L1 assessment beyond static, one-dimensional evaluation.

To improve predictive accuracy and facilitate personalized perioperative therapy, PD-L1 assessment is increasingly implemented across multiple modalities. Figure 3 illustrates a proposed workflow that combines CPS scoring, spatial profiling, multi-omics integration, microbiota analysis, and dynamic monitoring to guide adaptive immunotherapy strategies in GC[26,27]. This integrated approach enables context-aware, stage-specific evaluation, enhances patient stratification, and informs individualized perioperative treatment planning.

Figure 3 A proposed workflow that combines combined positive score evaluation, spatial profiling, multi-omics integration, microbiota analysis, and dynamic monitoring to guide adaptive immunotherapy strategies in gastric cancer. CPS: Combined positive score; IF: Immunofluorescence; TLS: Tertiary lymphoid structures; TMB: Tumor mutational burden; EBV: Epstein-Barr virus; MSI: Microsatellite instability.

Spatial biology

Tumor tissues exhibit substantial spatial heterogeneity in PD-L1 expression and immune cell distribution, which significantly influences tumor-immune dynamics and response to ICIs. Techniques such as spatial transcriptomics, multiplex immunofluorescence, and TLS mapping enable high-resolution visualization of the tumor microenvironment[25,26,55]. These approaches allow discrimination between immunologically active vs suppressive regions, providing a more nuanced understanding of PD-L1 function beyond bulk CPS scoring. From an author perspective, incorporating spatial profiling into biomarker assessment can reveal tumor regions amenable to targeted interventions and improve patient stratification for perioperative therapy[13,14].

Multi-omics integration

Multi-omics strategies, integrating transcriptomics, proteomics, epigenomics, and immune repertoire analysis, allow comprehensive characterization of both tumor and immune compartments[27,28,56]. By combining PD-L1 expression with molecular subtype, TMB, MSI, and immune microenvironmental features, multi-omics approaches generate context-aware biomarkers that reflect both tumor-intrinsic properties and functional immune status. Such integrated analyses can elucidate mechanisms of immune evasion, predict ICI responsiveness, and identify patients most likely to benefit from perioperative immunotherapy[23,31]. From an author perspective, these strategies should be incorporated into trial design to enhance precision biomarker-driven treatment.

Microbiota and immune modulation

Emerging evidence indicates that intratumoral and gut microbiota can modulate PD-L1 expression, immune infiltration, and ICI response[29,30,57]. Specific microbial communities, including EBV-positive tumors, Helicobacter pylori, Fusobacterium, and Akkermansia, influence local immune activity, tumor immunogenicity, and potentially treatment outcomes. Integrating microbiota profiling into PD-L1 assessment may reveal microbe-mediated immune mechanisms, providing new avenues for combination therapies and precision immunotherapy[27,28,58,59]. From an author perspective, consideration of microbial influences represents a critical, yet often underappreciated, dimension of context-aware PD-L1 evaluation.

Dynamic biomarker assessment

PD-L1 expression is inherently dynamic, influenced by neoadjuvant therapy, tumor evolution, and microenvironmental changes[15,60,61]. Serial biopsies and longitudinal monitoring capture therapy-induced modulation of PD-L1 and immune cell composition, enabling adaptive therapeutic strategies. Dynamic assessment identifies emerging resistance mechanisms, guides post-surgical adjuvant therapy, and optimizes timing and selection of perioperative ICIs[23,26]. From an author perspective, incorporating temporal profiling is essential to move biomarker evaluation beyond static measurements toward clinically actionable, adaptive decision-making.

Prospective multicenter validation

To ensure generalizability and reproducibility, prospective multicenter studies are critical for validating stage-aware PD-L1 interpretation and emerging biomarker strategies. Standardized protocols for tissue sampling, scoring, and multi-modal assessment minimize technical variability and allow reliable comparisons across centers[25,27,28]. Multi-institutional validation also facilitates regulatory acceptance and clinical implementation, establishing a framework for integrating advanced biomarker assessment into routine perioperative practice[13,14].

Summary

Integrating spatial biology, multi-omics profiling, microbiota analysis, dynamic monitoring, and multicenter validation represents the frontier of PD-L1 research in GC. These strategies enhance accuracy and reproducibility, facilitate context-aware patient stratification, and inform personalized perioperative immunotherapy. Collectively, they provide a roadmap for precision oncology, moving beyond conventional CPS evaluation toward a holistic, stage- and context-dependent understanding of tumor-immune interactions. From an author perspective, embracing these emerging directions is essential to translate PD-L1 biology into actionable clinical strategies that optimize patient outcomes.

AUTHOR PERSPECTIVES

The evidence reviewed above highlights the complex, stage-dependent nature of PD-L1 expression in GC, emphasizing the need for a nuanced, context-aware interpretation. Several key perspectives emerge, providing guidance for both clinical practice and future perioperative immunotherapy research.

Stage-aware interpretation is essential

PD-L1 expression cannot be universally interpreted without considering pathologic stage, immune microenvironment composition, and cellular origin[6,12,62]. Stage II and III tumors exhibit distinct immune landscapes, including variations in T-cell infiltration, myeloid suppressive populations, and tumor heterogeneity, all of which substantially influence PD-L1’s prognostic and predictive relevance. From an author perspective, adopting a stage-aware approach is critical to contextualize PD-L1 expression, reduce patient misclassification, and improve stratification accuracy in both clinical trials and routine perioperative care[25,31].

Context-dependent immunotherapy strategies

Therapeutic decisions should integrate PD-L1 expression with tumor stage, immune context, and complementary biomarkers such as MSI, EBV status, TMB, and spatial immune features, rather than relying solely on baseline CPS[14,31,63,64]. This context-dependent strategy supports adaptive perioperative approaches, maximizing efficacy while minimizing unnecessary exposure to ICIs in patients unlikely to benefit. Incorporating stage-specific immune profiles enables clinicians to tailor treatment intensity and timing, potentially improving both short-term and long-term outcomes in resectable and locally advanced GC[6,23].

Future research priorities

To advance precision immunotherapy, integration of emerging technologies is essential. Spatial biology techniques, including multiplex immunofluorescence and spatial transcriptomics, provide high-resolution mapping of PD-L1 and immune cell distribution[25,26]. Multi-omics profiling, including transcriptomics, proteomics, and epigenomics, enables comprehensive characterization of tumor-immune interactions[27,28]. Microbiota analysis, encompassing both intratumoral and gut microbial communities, offers additional insight into immune modulation[29,30]. Dynamic biomarker assessment, via serial sampling and longitudinal monitoring, allows evaluation of therapy-induced PD-L1 changes and adaptive treatment strategies[26]. Prospective multicenter validation remains critical for standardizing these methodologies and ensuring reproducibility across diverse patient populations[27]. From an author perspective, prioritizing these research directions will enhance both mechanistic understanding and clinical applicability of PD-L1 as a biomarker.

Summary

We advocate a paradigm shift from static, one-dimensional PD-L1 assessment to a stage- and context-integrated evaluation. This framework combines pathologic stage, immune microenvironment, molecular biomarkers, and temporal dynamics to provide precise, clinically actionable insights. Implementing such an approach enables perioperative immunotherapy in GC to be delivered with greater accuracy, optimizing patient selection, enhancing response rates, and ultimately improving survival outcomes.

CONCLUSION

PD-L1 expression in GC is strongly influenced by tumor stage, highlighting the necessity of stage-aware, context-dependent evaluation. Stage II and III tumors differ in tumor evolution, immune microenvironment composition, and T-cell functionality, which substantially affect PD-L1 interpretation for prognosis and immunotherapy responsiveness. Ongoing controversies, including limitations of CPS, ambiguity in cellular source, spatial heterogeneity, and dynamic changes under therapy, underscore the need for integrated, multi-dimensional biomarker strategies. Incorporating spatial profiling, multi-omics analyses, microbiota assessment, and longitudinal monitoring enables precise patient stratification and optimizes perioperative immunotherapy approaches.

In conclusion, a stage- and context-informed PD-L1 assessment provides a framework for personalized, adaptive perioperative immunotherapy, enhancing patient selection, therapeutic efficacy, and clinical outcomes while minimizing unnecessary treatment exposure. From an author perspective, adopting a dynamic, integrated PD-L1 evaluation framework is essential for translating biomarker insights into actionable clinical strategies that improve patient outcomes and guide future precision oncology research.