Synergistic value of systemic inflammatory and tumor markers in predicting pretreatment distant metastasis of gastric cancer

Abstract

BACKGROUND

Early identification of distant metastasis before treatment is crucial for optimizing therapeutic strategies in patients with gastric cancer. However, conventional imaging modalities are limited in detecting occult metastasis, highlighting the need for reliable, accessible biomarkers to improve pre-treatment risk stratification.

AIM

To develop and validate a predictive model based on the integration of inflammatory and tumor markers for pre-treatment prediction of distant metastasis in gastric cancer.

METHODS

A total of 279 patients with newly diagnosed gastric adenocarcinoma at the Affiliated Hospital of Xuzhou Medical University from January 2020 to December 2024 were retrospectively enrolled and randomly divided into a training set (n = 152) and a validation set (n = 127). Clinical characteristics, peripheral blood inflammatory parameters [neutrophils (NE), lymphocytes (LY), monocytes, platelets], derived inflammatory indices [neutrophil-to-lymphocyte ratio, lymphocyte-to-monocyte ratio, platelet-to-lymphocyte ratio, systemic immune-inflammation index (SII), and systemic inflammation response index (SIRI), and tumor markers (carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9)] were collected. The calculation formulas for the indices were as follows: SII = (NE × platelets)/LY, and SIRI = (NE × monocytes)/LY. Variables associated with distant metastasis were screened using univariate analysis and least absolute shrinkage and selection operator regression. A multivariate logistic regression model was then constructed and internally validated. Model performance was evaluated using the receiver operating characteristic curve, calibration curve, Brier score, and decision curve analysis.

RESULTS

Tumor location, CEA, CA19-9, neutrophil-to-lymphocyte ratio, lymphocyte-to-monocyte ratio, platelet-to-lymphocyte ratio, SII and SIRI were significantly associated with distant metastasis in univariate analysis (P < 0.05). Least absolute shrinkage and selection operator regression identified tumor location, CEA, CA19-9, NE, and LY as candidate predictors. Multivariate analysis revealed that CA19-9 [odds ratio (OR) = 1.019, P = 0.022] and NE (OR = 1.604, P = 0.005) were independent risk factors, whereas LY (OR = 0.244, P = 0.017) and tumor location in the gastric antrum (OR = 0.362, P = 0.036) were protective factors. The combined model achieved an area under the curve of 0.888 (95% confidence interval: 0.832-0.937) in the training set and 0.896 (95% confidence interval: 0.835-0.945) in the validation set. Calibration analysis demonstrated good agreement between predicted and observed outcomes, and decision curve analysis indicated favorable clinical net benefit across a wide range of threshold probabilities.

CONCLUSION

A predictive model incorporating CA19-9, NE, LY, and tumor location demonstrates good discrimination and calibration for predicting distant metastasis before treatment in gastric cancer. This practical and cost-effective model may assist clinicians in early risk stratification and individualized decision-making.

Key Words: Gastric cancer; Distant metastasis; Inflammatory indicators; Tumor markers; Predictive model; Calibration curve

Core Tip: Early and accurate identification of distant metastasis before treatment remains a clinical challenge in gastric cancer. This study developed and internally validated a practical predictive model integrating routinely available inflammatory markers and tumor markers. By combining carbohydrate antigen 19-9, neutrophil count, lymphocyte count, and tumor location, the model demonstrated good discrimination, calibration, and clinical net benefit. The proposed model provides a convenient and cost-effective tool for pre-treatment risk stratification, which may assist clinicians in optimizing diagnostic strategies and individualized management for patients with gastric cancer.

INTRODUCTION

Gastric cancer remains one of the most common malignant tumors worldwide and continues to pose a major public health burden. According to the GLOBOCAN 2020 estimates, gastric cancer ranks fifth in global cancer incidence and fourth in cancer-related mortality, with more than one million new cases and approximately 769000 deaths reported annually[1]. Despite advances in diagnostic techniques and therapeutic strategies, the prognosis of gastric cancer remains unsatisfactory, largely due to the high proportion of patients presenting with distant metastasis at initial diagnosis. It is estimated that 20%-30% of patients are diagnosed with metastatic disease at first presentation, for whom curative surgery is no longer feasible and systemic therapy becomes the mainstay of treatment[2].

Accurate identification of distant metastasis before treatment is therefore crucial for treatment planning and prognostic assessment in gastric cancer. Currently, imaging modalities such as computed tomography (CT), magnetic resonance imaging, and positron emission tomography-CT are considered the gold standard for metastatic evaluation. However, these techniques have inherent limitations. Small or occult metastatic lesions may escape detection, as most imaging methods are insensitive to lesions smaller than 5 mm[3]. In addition, imaging accuracy is highly dependent on equipment availability and operator expertise, which restricts its widespread application, particularly in primary medical institutions. Consequently, there is an unmet clinical need for convenient, objective, and reproducible biomarkers that can assist in the early identification of distant metastasis before treatment.

Cancer progression and metastasis are closely associated with systemic inflammation and immune dysregulation[4,5]. Increasing evidence suggests that inflammatory responses can promote tumor invasion and metastasis by facilitating angiogenesis, remodeling the extracellular matrix, and inducing immunosuppression within the tumor microenvironment[6,7]. Peripheral blood immune cells, such as neutrophils, lymphocytes, monocytes, and platelets, reflect the host’s systemic inflammatory and immune status. Derived inflammatory indices, including the neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), and platelet-to-lymphocyte ratio (PLR), have been widely reported as prognostic or predictive indicators in various malignancies, owing to their simplicity, low cost, and accessibility[8,9].

Tumor markers also play an important role in the clinical management of gastric cancer. Carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9) are routinely measured in clinical practice and have been associated with tumor burden, disease progression, and metastatic potential. Elevated pre-treatment levels of these markers have been linked to advanced disease stage and poor prognosis in gastric cancer[10]. However, most existing studies have focused on either inflammatory markers or tumor markers alone, and the predictive performance of single indicators remains limited.

To date, there is a lack of comprehensive models that systematically integrate inflammatory markers and tumor markers for the pre-treatment prediction of distant metastasis in gastric cancer. Moreover, the generalizability and clinical utility of such models have not been sufficiently validated. Therefore, this study aimed to analyze pre-treatment inflammatory markers and tumor markers in patients with newly diagnosed gastric adenocarcinoma, identify independent risk factors for distant metastasis, and construct a combined predictive model. By incorporating routinely available clinical and laboratory parameters, this model seeks to provide a practical and reliable tool for early risk stratification and individualized clinical decision-making in gastric cancer.

MATERIALS AND METHODS

Study population and data collection

This retrospective study enrolled 279 consecutive patients with gastric adenocarcinoma who were initially diagnosed at the Affiliated Hospital of Xuzhou Medical University from January 2020 to December 2024. All patients were pathologically confirmed to have gastric adenocarcinoma and had not received any anti-tumor treatment prior to diagnosis. Using computer-generated simple random sampling, patients were randomly divided into a training set (n = 152) and a validation set (n = 127). The training set was used for variable screening and model construction, while the validation set served for internal validation.

Clinical data were extracted from electronic medical records, including sex, age, clinical stage, history of hypertension, history of diabetes mellitus, and primary tumor location (cardia, gastric body, or antrum). Laboratory data were obtained from routine blood tests performed within one week prior to initial treatment. Peripheral blood parameters included absolute neutrophil count (NE), lymphocyte count (LY), monocyte count, and platelet count. Based on these parameters, derived inflammatory indices were calculated, including the NLR, LMR, PLR, systemic immune-inflammation index (SII), and systemic inflammation response index (SIRI). Tumor markers, including CEA and CA19-9, were also collected. The calculation formulas for the indices were as follows: SII = (NE × platelet count)/LY, and SIRI = (NE × monocyte count)/LY.

Distant metastasis was defined according to the 8^th edition of the American Joint Committee on Cancer Staging Manual and referred to metastasis to extra-gastric organs (such as the liver, lung, bone, or brain) or distant lymph nodes[11]. Metastatic status was confirmed by imaging modalities, including CT, magnetic resonance imaging, or positron emission tomography-CT, and/or pathological examination when available. This study was approved by the Institutional Ethics Committee (approval No. XYFY2025-KL581-01), and patient confidentiality was strictly maintained throughout the study.

Inclusion and exclusion criteria

Inclusion criteria: (1) Histopathologically confirmed gastric adenocarcinoma; (2) No prior anti-tumor treatment, including surgery, chemotherapy, radiotherapy, or immunotherapy, before initial diagnosis; (3) Complete clinical and laboratory data available at baseline; and (4) No history of other malignant tumors.

Exclusion criteria: (1) Presence of severe complications such as gastrointestinal obstruction, perforation, or major bleeding; (2) Acute infectious diseases or chronic inflammatory or autoimmune disorders that could influence systemic inflammatory markers; (3) Severe dysfunction of major organs, including the heart, liver, kidneys, or brain; and (4) Use of medications known to affect hematological or immune parameters, such as antiplatelet agents, glucocorticoids, or immunosuppressants, within one month prior to diagnosis.

Statistical analysis

Statistical analyses were performed using SPSS software (version 27.0), while Python and R 4.5.1 software were used for data visualization and model construction. Continuous variables were assessed for normality using the Shapiro-Wilk test. Normally distributed variables are presented as the mean ± SD, whereas non-normally distributed variables are expressed as the median and interquartile range. Categorical variables are reported as n (%).

Comparisons between groups were conducted using the independent samples Student’s t-test or the Mann-Whitney U test for continuous variables, as appropriate, and the χ² test for categorical variables. In the training set, univariate analyses were first performed to identify variables associated with distant metastasis. Variables with statistical significance were subsequently subjected to least absolute shrinkage and selection operator (LASSO) regression to reduce multicollinearity and minimize the risk of model overfitting.

Variables selected by LASSO regression were then incorporated into a multivariate logistic regression model to identify independent predictors of distant metastasis. A nomogram was constructed based on the final multivariate model. The discriminative ability of the model was evaluated using receiver operating characteristic (ROC) curves, and the area under the curve (AUC) with corresponding 95% confidence intervals (CIs) was calculated. For ROC analysis of categorical predictors, binary variables were coded numerically. The optimal cut-off value was determined based on the maximum Youden index to maximize the combined sensitivity and specificity.

Model calibration was assessed using calibration curves, the Hosmer-Lemeshow goodness-of-fit test, and the Brier score, with a P value greater than 0.05 indicating acceptable calibration. Decision curve analysis was performed to evaluate the clinical net benefit of the predictive model across a range of threshold probabilities. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics of the study population

Baseline characteristics were well-balanced between the training and validation sets, ensuring the comparability of the two groups (Table 1).

Table 1 Comparison of clinical characteristics between the training and validation sets, n (%).

Characteristic	Training set (n = 152)	Validation set (n = 127)	P value
Sex			0.774
Male	103 (67.8)	84 (66.1)
Female	49 (32.2)	43 (33.9)
Age			0.205
< 60 years	71 (46.7)	69 (53.4)
≥ 60 years	81 (53.3)	58 (45.7)
Stage			0.823
I-II	65 (42.8)	56 (44.1)
III-IV	87 (57.2)	71 (55.9)
Hypertension			0.179
No	113 (74.3)	103 (81.1)
Yes	39 (25.7)	24 (18.9)
Diabetes			0.543
No	129 (84.9)	111 (87.4)
Yes	23 (15.1)	16 (12.6)
Tumor location			0.788
Cardia	51 (33.6)	38 (29.9)
Antrum	78 (51.3)	70 (55.1)
Body	23 (15.1)	19 (15.0)
Distant metastasis			0.942
No	94 (61.8)	78 (61.4)
Yes	58 (38.2)	49 (38.6)

Baseline and laboratory characteristics in the training set: Metastatic vs non-metastatic groups

The results of the Shapiro-Wilk test indicated that all laboratory indicators, significantly deviated from a normal distribution (all P < 0.05). In the training set, no statistically significant differences were observed between the metastatic and non-metastatic groups with respect to sex or age (P > 0.05). Tumor location differed significantly between the two groups (P < 0.05). Regarding laboratory indicators, the metastatic group exhibited higher levels of CEA, CA19-9, NLR, PLR, SII and SIRI compared with the non-metastatic group, whereas LMR was lower (all P < 0.05) (Table 2). These findings indicate that tumor location and several pre-treatment inflammatory/tumor-marker-related indices are associated with distant metastasis in the training cohort.

Table 2 Comparison of gastric cancer patients between metastatic and non-metastatic groups in the training set, n (%)/median (interquartile range).

Characteristic	Non-metastatic group (n = 94)	Metastatic group (n = 58)	P value
Sex			0.914
Male	64 (68.1)	39 (67.2)
Female	30 (31.9)	19 (32.8)
Age			0.426
< 60 years	44 (46.8)	27 (46.6)
≥ 60 years	50 (53.2)	31 (53.4)
Tumor location			< 0.001
Cardia	22 (23.4)	29 (50.0)
Antrum	59 (62.8)	19 (32.8)
Body	13 (13.8)	10 (17.2)
CEA	2.25 (1.56, 3.98)	7.42 (1.92, 79.93)	< 0.001
CA19-9	11.31 (6.04, 18.19)	57.45 (7.91, 175.68)	< 0.001
NE	2.91 (2.38, 3.79)	4.06 (2.84, 6.83)	< 0.001
LY	1.50 (1.20, 1.90)	1.20 (0.90, 1.50)	< 0.001
MON	0.36 (0.29, 0.44)	0.41 (0.32, 0.50)	0.105
PLT	220.50 (187.50, 270.50)	223.50 (181.25, 300.50)	0.799
NLR	1.95 (1.33, 2.76)	3.57 (2.77, 5.55)	< 0.001
LMR	4.44 (3.07, 6.15)	2.86 (2.30, 3.85)	< 0.001
PLR	150.24 (105.08, 212.00)	186.00 (130.61, 302.65)	0.002
SII	439.54 (298.28, 681.39)	906.07 (505.75, 1785.25)	< 0.001
SIRI	0.76 (0.43, 0.76)	1.48 (0.84, 2.94)	< 0.001

Calculation methods: Neutrophil-to-lymphocyte ratio = neutrophils/Lymphocytes; lymphocyte-to-monocyte ratio = lymphocytes/monocytes; platelet-to-lymphocyte ratio = platelets/Lymphocytes; systemic immune-inflammation index = (neutrophils × platelets)/Lymphocytes; systemic inflammation response index = (neutrophils × monocytes)/Lymphocytes. CEA: Carcinoembryonic antigen; CA19-9: Carbohydrate antigen 19-9; NE: Neutrophils; LY: Lymphocytes; MON: Monocytes; PLT: Platelets; NLR: Neutrophil-to-lymphocyte ratio; LMR: Lymphocyte-to-monocyte ratio; PLR: Platelet-to-lymphocyte ratio; SII: Systemic immune-inflammation index; SIRI: Systemic inflammation response index.

Identification of independent factors associated with distant metastasis in the training set

Variables showing statistical significance in univariate analysis (P < 0.05) were further subjected to LASSO regression for feature selection. Tumor location (antrum), CEA, CA19-9, NE, LY, and LMR were subsequently included in the multivariate logistic regression model. Multivariate analysis demonstrated that CA19-9 [odds ratio (OR) = 1.019, P = 0.022] and NE (OR = 1.604, P = 0.005) were independently associated with an increased risk of distant metastasis, whereas LY (OR = 0.244, P = 0.017) and tumor location in the gastric antrum (OR = 0.362, P = 0.036) were independently associated with a decreased risk of distant metastasis (Table 3). CEA and LMR did not retain independent significance in the multivariate model.

Table 3 Multivariate logistic regression analysis in the training set.

	B	Wald χ2	P value	OR	95% confidence interval
Antrum	-1.015	5.913	0.036	0.362	(0.141-0.935)
CEA	0.003	1.164	0.410	1.003	(0.996-1.009)
CA19-9	0.019	8.140	0.022	1.019	(1.003-1.036)
NE	0.473	5.470	0.005	1.604	(1.151-2.236)
LY	-1.412	4.191	0.017	0.244	(0.077-0.774)
LMR	-0.008	1.635	0.954	0.992	(0.754-1.305)

Calculation methods: Lymphocyte-to-monocyte ratio = lymphocytes/monocytes. CEA: Carcinoembryonic antigen; CA19-9: Carbohydrate antigen 19-9; NE: Neutrophils; LY: Lymphocytes; LMR: Lymphocyte-to-monocyte ratio.

Development and validation of the combined predictive model for distant metastasis

A combined predictive model was constructed based on the independent predictors identified in multivariate analysis - CA19-9, NE, LY, and tumor location in the gastric antrum - and a corresponding nomogram was developed. Model discrimination was evaluated using ROC curve analysis. For ROC analysis of tumor location, the variable was dichotomized as antrum vs non-antrum (cardia/body), and the corresponding AUC was calculated based on this binary classification. The optimal cut-off value was defined as the point that maximized the Youden index, and the sensitivity and specificity of each single indicator as well as the combined model were calculated accordingly (Table 4; Figures 1 and 2).

Figure 1 Nomogram model for predicting the risk of distant metastasis in patients with gastric adenocarcinoma. CA19-9: Carbohydrate antigen 19-9; NE: Neutrophils; LY: Lymphocytes.

Figure 2 Receiver operating characteristic curves of the nomogram model for predicting the risk of metastasis in patients with gastric adenocarcinoma. A: Training set; B: Validation set. ROC: Receiver operating characteristic curves; AUC: Area under the curve.

Table 4 Performance of the combined predictive model and individual indicators for predicting distant metastasis in gastric cancer.

	Cut-off	AUC	Sensitivity	Specificity	Youden index
Training set
CA19-9	22.9	0.732	0.638	0.872	0.510
NE	4.75	0.731	0.448	0.947	0.395
LY	1.4	0.687	0.672	0.606	0.278
Antrum	-	0.651	0.672	0.628	0.300
Model	0.337	0.888	0.776	0.819	0.595
Validation set
CA19-9	8.49	0.696	0.714	0.590	0.308
NE	6.09	0.781	0.633	0.872	0.504
LY	1.3	0.721	0.694	0.679	0.627
Antrum	-	0.683	0.674	0.693	0.373
Model	0.311	0.896	0.898	0.718	0.616

CA19-9: Carbohydrate antigen 19-9; NE: Neutrophils; LY: Lymphocytes.

In the training set, the combined model achieved an AUC of 0.888 (95%CI: 0.832-0.937), while in the validation set the AUC was 0.896 (95%CI: 0.835-0.945), indicating good discriminative performance in both the training and validation sets. Calibration in the training set demonstrated good agreement between predicted probabilities and observed outcomes, with a Hosmer-Lemeshow test χ² = 2.782 (P > 0.05) and a Brier score of 0.126 (Figure 3). Decision curve analysis showed that, across a wide range of threshold probabilities, the combined model provided a higher net benefit than the “treat-all” and “treat-none” strategies in both the training and validation sets (Figure 4), supporting its potential clinical utility.

Figure 3 Calibration curves comparing the predicted risk of metastasis by the nomogram model with the actual incidence in patients with gastric adenocarcinoma. A: Training set; B: Validation set.

Figure 4  Decision curve analysis of the nomogram model for predicting the risk of metastasis and the actual incidence in patients with gastric adenocarcinoma.

DISCUSSION

Gastric cancer remains a highly prevalent malignancy of the digestive system, with gastric adenocarcinoma representing its most common histological subtype[12]. The presence of distant metastasis at the time of initial diagnosis is a major determinant of poor prognosis and treatment failure. Given the inherent limitations of conventional imaging in detecting occult or early distant metastasis, developing accessible and cost-effective pre-treatment predictive tools for gastric cancer remains a critical clinical need[13,14].

In the present study, we retrospectively analyzed pre-treatment inflammatory markers and tumor markers in patients with newly diagnosed gastric adenocarcinoma and identified CA19-9, NE, LY, and tumor location as independent factors associated with distant metastasis. Based on these variables, a combined predictive model was constructed and internally validated. The model demonstrated good discriminative ability in both the training and validation sets, as reflected by AUC values of 0.888 and 0.896, respectively. Calibration analysis further indicated satisfactory agreement between predicted probabilities and observed outcomes, while decision curve analysis suggested potential clinical benefit across a wide range of threshold probabilities. In recent years, various predictive models have been developed to evaluate the risk of metastasis in gastric cancer. For instance, a study reported that CEA yielded an AUC of only 0.694 for predicting metastasis-related features in advanced gastric cancer, showing significantly lower accuracy than our integrated approach[15]. By comparison, our model achieved a comparable predictive efficacy using only four routinely available clinical parameters. Collectively, these findings indicate that the proposed model may serve as a useful adjunct to conventional imaging for pre-treatment risk stratification.

CA19-9 is a widely used tumor marker associated with several adenocarcinomas and is commonly considered a surrogate indicator of tumor burden and biological aggressiveness[10]. Elevated CA19-9 levels have been reported to correlate with advanced disease stage, increased metastatic potential, and unfavorable prognosis in gastric cancer[16,17]. Mechanistically, carbohydrate antigens expressed on tumor cell surfaces may interact with selectins on vascular endothelial cells, facilitating tumor cell adhesion, extravasation, and subsequent colonization of distant organs[18]. Consistent with these observations, the present study found that pre-treatment CA19-9 levels were significantly higher in patients with distant metastasis and remained an independent risk factor in multivariate analysis, supporting its role as a key component of metastasis-related risk assessment.

Systemic inflammatory status also plays an essential role in tumor progression and metastasis. Neutrophils and lymphocytes represent two critical components of the host immune response, and the imbalance between pro-tumor inflammation and anti-tumor immunity can contribute to metastatic dissemination[19]. Elevated neutrophils contribute to the pre-metastatic niche by secreting pro-angiogenic factors such as vascular endothelial growth factor and matrix metalloproteinases, which promote tumor cell extravasation and angiogenesis[20,21]. Furthermore, neutrophils are capable of forming neutrophil extracellular traps. These net-like structures, composed of DNA, histones, and other components, can capture circulating tumor cells and may facilitate their immune evasion[22]. Conversely, a reduction in lymphocytes - particularly CD8+ T cells and natural killer cells - suggests that the body’s antitumor immune surveillance may have been compromised. In this environment, tumor cells are able to evade cytotoxic killing and successfully metastasize to and establish themselves in distant organs[23]. In this study, increased NEs and decreased LYs were independently associated with distant metastasis, suggesting that patients with metastatic disease exhibit a more pronounced inflammatory response accompanied by immune suppression. These findings are in line with previous reports highlighting the prognostic and predictive value of inflammation-related biomarkers in solid tumors[24]. Notably, the present study utilized the absolute values of NE and LY, as absolute counts preserve the independent predictive power of each cell type and are less susceptible to the mathematical fluctuations inherent in ratio calculations. Furthermore, this approach provides more direct clinical thresholds for risk stratification and allows for a more granular assessment of the host’s immune-inflammatory status prior to treatment.

In addition to laboratory indicators, primary tumor location emerged as an independent factor associated with distant metastasis. Tumors located in the gastric antrum were less likely to present with distant metastasis compared with those arising in the cardia or gastric body. This observation may be partially explained by anatomical and biological differences among gastric subsites. The cardia is characterized by a rich lymphatic network that may facilitate early lymphatic spread and subsequent systemic dissemination, whereas tumors of the gastric body often display more heterogeneous biological behavior[25]. In contrast, antral tumors tend to exhibit relatively localized growth patterns, with distant metastasis occurring at a later stage[26]. Beyond the influence of anatomical and lymphatic drainage characteristics, the observed differences are potentially driven by the fundamental molecular heterogeneity across gastric regions. Gastric cancer is a highly molecularly diverse disease with subtypes such as Epstein-Barr virus-positive, microsatellite unstable, and chromosomal instability, which often exhibit distinct spatial preferences[27]. For instance, chromosomal instability subtypes are more frequent in the cardia and gastroesophageal junction, whereas microsatellite unstable status and certain inflammatory-related markers vary significantly across the antrum and body. These molecular differences likely influence the distinct metastatic potential of tumors from different locations[28]. Although the underlying mechanisms require further investigation, tumor location appears to be a clinically relevant factor that should be considered in metastasis risk assessment.

Given the multifactorial nature of tumor metastasis, reliance on a single biomarker is unlikely to provide sufficient predictive accuracy[29,30]. By integrating tumor burden (CA19-9), systemic inflammatory status (neutrophil and LYs), and a key clinical characteristic (tumor location), the combined model developed in this study offers a more comprehensive reflection of the tumor-host interaction. The favorable performance of the model in both discrimination and calibration suggests a certain degree of robustness and generalizability. Importantly, all variables included in the model are routinely available in clinical practice, which enhances its feasibility and potential applicability.

Nevertheless, several limitations of this study should be acknowledged. First, the retrospective design and single-center setting may introduce selection bias. Second, the sample size was relatively limited, and certain molecular or pathological features, such as human epidermal growth factor receptor 2 status, were not incorporated into the analysis. Third, although internal validation was performed, further validation in larger, multicenter, and prospective cohorts is warranted. Future large-scale, multicenter, prospective studies that incorporate comprehensive molecular classifications are essential to rigorously validate and further refine this predictive model.

In conclusion, this study developed and validated a practical predictive model based on pre-treatment inflammatory markers, tumor markers, and tumor location for the prediction of distant metastasis in gastric cancer. The model demonstrates good discriminative performance and potential clinical value, providing a convenient tool for early risk stratification and supporting individualized decision-making in the management of gastric cancer.

CONCLUSION

In summary, this study developed and validated a practical predictive model integrating pre-treatment inflammatory markers, tumor markers, and tumor location to estimate the risk of distant metastasis in gastric cancer. The model demonstrated good discriminative ability and calibration, suggesting its potential value as an adjunctive tool for early risk stratification. By relying on routinely available clinical parameters, this approach may support individualized diagnostic and therapeutic decision-making in patients with gastric cancer.