Multidimensional integration: A novel breakthrough in prognostic prediction for immunochemotherapy in human epidermal growth factor receptor-2-negative advanced gastric cancer

Abstract

The combination of programmed cell death-1 inhibitors with chemotherapy has significantly improved clinical outcomes in patients with human epidermal growth factor receptor-2-negative advanced gastric cancer. However, substantial interindividual heterogeneity in treatment response limits the precision of clinical decision-making. Recent efforts have focused on developing prognostic models that integrate multidimensional data, including molecular biomarkers, clinicopathological characteristics, and systemic inflammatory-nutritional indices. Among these, nomogram-based models have shown promise in improving individualized risk stratification. This review summarizes current advances in immunochemotherapy for advanced gastric cancer, highlights the emerging role of multidimensional prognostic modeling, discusses existing controversies and limitations, and provides perspectives on future directions, including the integration of liquid biopsy and artificial intelligence. A more comprehensive and dynamic predictive framework is essential to optimize treatment selection and improve patient outcomes.

Key Words: Gastric cancer; Immunochemotherapy; Prognostic prediction; Multidimensional nomogram; Efficacy

Core Tip: Multidimensional prognostic models integrating molecular, clinical, and host-related factors represent a promising approach for predicting outcomes in human epidermal growth factor receptor-2 negative advanced gastric cancer undergoing immunochemotherapy. Despite encouraging progress, challenges remain regarding model generalizability, biomarker standardization, and clinical applicability. Future integration of multi-omics data and artificial intelligence may enable more precise and dynamic prediction systems.

INTRODUCTION

Gastric cancer remains one of the leading causes of cancer-related mortality worldwide, with advanced-stage disease accounting for the majority of deaths[1]. Among these, human epidermal growth factor receptor-2-negative advanced gastric cancer represents a predominant clinical subtype with limited therapeutic options and poor prognosis[2-4]. Although conventional chemotherapy has historically been the standard treatment, survival outcomes remain unsatisfactory[5].

In recent years, the advent of immunotherapy, particularly immune checkpoint inhibitors targeting programmed cell death-1 (PD-1) and programmed death-ligand 1 (PD-L1), has revolutionized the treatment landscape[6,7]. The combination of PD-1 inhibitors with chemotherapy has demonstrated significant survival benefits and has been established as a first-line treatment strategy for advanced gastric cancer[8-10]. However, only a subset of patients derives durable benefit from immunochemotherapy, highlighting the urgent need for reliable prognostic and predictive tools[11,12].

In this context, multidimensional prognostic modeling has emerged as a promising approach[13]. By integrating diverse biological and clinical information, these models aim to capture tumor heterogeneity and host-related factors more comprehensively than traditional single-parameter predictors[14,15].

CURRENT ADVANCES IN IMMUNOCHEMOTHERAPY FOR ADVANCED GASTRIC CANCER

Immune checkpoint inhibitors combined with chemotherapy

Large-scale randomized clinical trials have established the efficacy of combining PD-1 inhibitors with chemotherapy in advanced gastric cancer[9,12,16]. Studies such as CheckMate 649[9] and ORIENT-16[12] have demonstrated significant improvements in overall survival and progression-free survival compared with chemotherapy alone, particularly in patients with higher PD-L1 expression. These findings have led to the incorporation of immunochemotherapy into standard first-line treatment regimens[17]. Nevertheless, therapeutic responses vary widely among patients, reflecting the complex interplay between tumor biology and host immunity[18].

Biomarkers for response prediction

Several biomarkers have been explored to predict response to immunochemotherapy. PD-L1 expression remains the most widely used biomarker in clinical practice, although its predictive value is inconsistent across studies[19-21]. Microsatellite instability-high status is associated with enhanced immunotherapy responsiveness but is present in only a small proportion of patients[22,23]. Tumor mutational burden has also been investigated but lacks standardized thresholds and clinical applicability in gastric cancer[24,25]. Overall, reliance on single biomarkers has proven insufficient to fully capture the complexity of treatment response, prompting the development of integrative predictive approaches[26].

MULTIDIMENSIONAL PROGNOSTIC MODELS IN GASTRIC CANCER

Multidimensional prognostic models integrate diverse categories of variables, including molecular, clinicopathological, inflammatory, nutritional, and computational features, as summarized in Table 1. A conceptual framework of multidimensional prognostic modeling integrating molecular, clinical, and host-related factors is illustrated in Figure 1.

Figure 1 Multidimensional framework for prognostic prediction in human epidermal growth factor receptor-2-negative advanced gastric cancer undergoing immunochemotherapy. PD-L1: Programmed cell death ligand 1; MSI: Microsatellite instability; MMR: Mismatch repair; EBV: Epstein-Barr virus; ctDNA: Circulating tumor DNA; TNM: Tumor-node-metastasis; ECOG: Eastern Cooperative Oncology Group; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; LMR: Lymphocyte-to-monocyte ratio; PNI: Prognostic nutritional index; CALLY: C-reactive protein-albumin-lymphocyte; CRP: C-reactive protein; PD-1: Programmed cell death protein 1; AI: Artificial intelligence.

Table 1 Representative biomarkers and modeling dimensions for prognostic prediction in human epidermal growth factor receptor-2-negative advanced gastric cancer.

Dimension	Representative biomarkers/features	Biological/clinical relevance	Current limitations
Molecular	PD-L1, MSI, TMB, EBV, ctDNA	Reflect tumor immunogenicity and predict response to immune checkpoint inhibitors	Heterogeneity of assays; lack of standardized thresholds
Clinicopathological	TNM stage, histological differentiation, ECOG performance status	Indicate tumor burden, disease stage, and patient functional status	Static variables; limited ability to capture dynamic tumor evolution
Inflammatory	NLR, PLR, LMR, SII	Reflect systemic inflammatory response associated with tumor progression	Variability in cutoff values; influenced by non-cancer conditions
Nutritional	Albumin, PNI, CALLY index	Reflect host nutritional and immunological status	Susceptible to comorbidities and acute clinical conditions
Computational modeling	Nomogram, machine learning, radiomics	Enable individualized risk prediction through integration of multidimensional data	Limited external validation; risk of overfitting; lack of clinical interpretability

PD-L1: Programmed cell death ligand 1; MSI: Microsatellite instability; TMB: Tumor mutational burden; EBV: Epstein-Barr virus; ctDNA: Circulating tumor DNA; TNM: Tumor-node-metastasis; ECOG: Eastern Cooperative Oncology Group; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; LMR: Lymphocyte-to-monocyte ratio; SII: Systemic Immune-Inflammation Index; PNI: Prognostic nutritional index; CALLY: C-reactive protein-albumin-lymphocyte.

Limitations of single-parameter prediction

Traditional prognostic models based on isolated molecular or clinical factors fail to account for the multifactorial nature of tumor progression and therapeutic response[27]. The heterogeneity of gastric cancer necessitates a more comprehensive framework that integrates multiple dimensions of patient data[28].

Emergence of integrative modeling approaches

Recent years have witnessed a shift toward multidimensional prognostic modeling, which incorporates molecular biomarkers, clinicopathological variables, and systemic host-related indicators[29]. This approach reflects the recognition that tumor behavior is influenced not only by intrinsic genetic alterations but also by the tumor microenvironment and host systemic status[30].

Representative multidimensional model

A notable example of this paradigm is the nomogram developed by Yao et al[31], which integrates immune-related biomarkers (such as PD-L1 expression and microsatellite status), clinicopathological features (including tumor-node-metastasis stage and tumor differentiation), and inflammatory-nutritional indices such as the neutrophil-to-lymphocyte ratio and C-reactive protein-albumin-lymphocyte index.

This model demonstrates improved predictive performance by capturing multiple aspects of tumor biology and host response[31]. Furthermore, it provides individualized risk stratification for both short-term progression and long-term survival, thereby supporting clinical decision-making[32,33]. From a practical perspective, such models may help identify patients who are more likely to benefit from immunochemotherapy and guide treatment optimization[34]. However, their clinical implementation requires further validation.

CONTROVERSIES AND CHALLENGES

Despite encouraging progress, several challenges remain in the development and application of multidimensional prognostic models[35]. First, the lack of external validation limits the generalizability of many models. Most studies are based on retrospective single-center cohorts with relatively small sample sizes, which may introduce bias[36]. Second, variability in biomarker detection methods, particularly for PD-L1 expression, complicates cross-study comparisons and clinical standardization[37]. Third, the integration of heterogeneous data types presents methodological challenges, including data harmonization and model overfitting[38]. Finally, most existing models are static and fail to capture the dynamic evolution of tumor biology during treatment[39,40].

AUTHOR’S PERSPECTIVE

Multidimensional integration represents a critical direction for the future of prognostic prediction in gastric cancer. However, current models remain in an early stage of development. We believe that the next generation of predictive systems should move beyond static nomograms toward dynamic, adaptive models that incorporate longitudinal data.

In addition, the incorporation of emerging biomarkers, such as circulating tumor DNA and immune profiling, may further enhance predictive accuracy[41]. Importantly, model development should prioritize clinical applicability, ensuring that predictive tools are both robust and easily implementable in routine practice[42-44]. The current challenges and potential future directions of prognostic modeling in immunochemotherapy are summarized in Figure 2.

Figure 2 Current challenges and future directions of prognostic modeling in immunochemotherapy for advanced gastric cancer. PD-L1: Programmed cell death ligand 1; IHC: Immunohistochemistry; AI: Artificial intelligence; ctDNA: Circulating tumor DNA; ML: Machine learning.

FUTURE DIRECTIONS

Future research should focus on several key areas. Large-scale, multicenter prospective studies are essential to validate existing models and improve their generalizability[45]. In addition, external validation using real-world data from different geographic regions and clinical settings will be critical to ensure robustness and reproducibility[46]. Standardization of biomarker assessment methods, particularly for PD-L1 expression and emerging molecular markers, is also necessary to facilitate cross-study comparability and clinical implementation[47].

Second, the integration of liquid biopsy techniques, including circulating tumor DNA, circulating tumor cells, and other blood-based biomarkers, offers a minimally invasive approach for real-time monitoring of tumor dynamics and treatment response[48-50]. These technologies may enable longitudinal risk assessment and early detection of therapeutic resistance, thereby supporting adaptive treatment strategies[51,52].

Third, advances in artificial intelligence and machine learning are expected to play a central role in the development of next-generation predictive models[53,54]. Deep learning and multimodal data integration approaches can effectively handle high-dimensional datasets, including genomics, transcriptomics, radiomics, and clinical variables, thereby improving predictive performance and uncovering complex nonlinear relationships[55-57]. However, improving model interpretability and ensuring transparency will be essential for clinical adoption.

Fourth, future models should evolve from static prediction tools toward dynamic and adaptive systems that incorporate longitudinal data[58,59]. Such models can capture temporal changes in tumor biology and host status during treatment, enabling continuous risk recalibration and personalized therapeutic decision-making. Ultimately, the combination of multi-omics data, imaging, and clinical information may enable the construction of comprehensive and dynamic prognostic systems[60].

CONCLUSION

Multidimensional prognostic modeling represents a promising strategy for improving individualized prediction in human epidermal growth factor receptor-2-negative advanced gastric cancer undergoing immunochemotherapy. By integrating diverse sources of information, these models offer a more comprehensive understanding of treatment response. However, significant challenges remain, and further research is needed to translate these models into routine clinical practice.