Abstract
Purpose
The management of advanced gastric cancer (GC) follows a binary paradigm: surgery for localized disease versus palliative care for metastatic disease. Oligometastatic gastric cancer (OMGC) represents an intermediate state in which the role of surgery remains uncertain. Because current definitions of OMGC are based primarily on anatomic criteria and do not capture underlying biological heterogeneity, we developed the Tumor-Induced Perturbation score (TIPscore) to quantify systemic tumor–host perturbation and support biology-informed risk stratification.
Materials and Methods
Patients with advanced GC were stratified into locally advanced gastric cancer (LAGC), OMGC, and widely metastatic gastric cancer (WMGC) groups. Baseline serum samples from 107 patients were used to develop the TIPscore in the discovery cohort (LAGC=55, OMGC=22, WMGC=30), and post-treatment serum samples from 104 patients were used for longitudinal validation (LAGC=71, OMGC=30, WMGC=3). An independent temporal cohort of 147 patients with baseline serum samples was used for external validation (LAGC=105, OMGC=28, WMGC=14). Serum proteomic profiling was performed using the Olink platform, and the TIPscore was constructed by Elastic Net regression.
Results
Three core proteins (IL-22RA1, HGF, and 4E-BP1) were incorporated into the TIPscore. The model positioned patients on a biological continuum of metastatic disease and showed consistent discriminatory performance across cohorts. The post-treatment TIPscore predicted 1-year overall survival in OMGC and may help stratify risk among patients with OMGC being considered for curative-intent surgery.
Conclusions
The TIPscore provides a serum-based, biologically grounded framework for refining OMGC stratification and informing surgical decision-making.
Trial Registration
Chinese Clinical Trial Registry: ChiCTR2500096665
Keywords: Gastric cancer, Metastasis, Liquid biopsy, Precision medicine, Surgical oncology
INTRODUCTION
Gastric cancer (GC) is one of the most common malignancies worldwide [1], with a substantial proportion of patients presenting at an advanced stage [2]. Management of advanced GC has long followed a binary paradigm: curative-intent surgery for localized disease and systemic therapy for metastatic disease [3,4]. However, recognition of oligometastatic gastric cancer (OMGC) has challenged this dichotomy because a subset of patients with limited metastatic burden may still benefit from surgical intervention [5,6,7,8]. In line with this clinical shift, the Chinese Society of Clinical Oncology and Japanese Gastric Cancer Association guidelines conditionally recommend conversion surgery for OMGC [9,10].
Despite increasing clinical interest in conversion surgery for OMGC, outcomes remain inconsistent. Although several nonrandomized studies have suggested survival benefits [11,12], randomized evidence has failed to demonstrate a clear advantage [13,14]. This discrepancy highlights a fundamental limitation of current clinical practice: surgical decision-making remains largely empirical and based on anatomic assessment. Suboptimal patient selection may therefore contribute to the variability in observed outcomes rather than indicate a true lack of surgical efficacy.
This challenge in patient selection stems largely from the current definition of OMGC. OMGC is defined according to the distribution of metastases rather than the underlying biology of the disease [6,7,9,10,11,12,13,15]. Metastatic progression represents a continuous biological process shaped by tumor–host interactions [16], and this process cannot be fully captured by anatomic staging alone. Consequently, patients classified within the same anatomic category may exhibit markedly different clinical outcomes.
We therefore hypothesized that OMGC should be redefined as a biological state rather than a purely anatomic entity. To address this, we developed a liquid biopsy–based model, the Tumor-Induced Perturbation score (TIPscore), which quantifies systemic perturbation and situates patients along a biological continuum of metastatic disease. We further evaluated its clinical utility for risk stratifying patients being considered for surgery and predicting treatment outcomes.
MATERIALS AND METHODS
Study population and clinical classification
Patients with advanced gastric or gastroesophageal junction adenocarcinoma were identified in a prospectively maintained database at Zhongshan Hospital, Fudan University. Among patients treated between December 2014 and July 2024, 107 patients with available baseline serum samples were included in the discovery cohort, and 104 patients with post-treatment serum samples available before surgery were included in the longitudinal validation cohort. The discovery cohort was used to construct a baseline serum-based biological staging model, whereas the longitudinal validation cohort was used to evaluate the dynamic and prognostic value of the post-treatment TIPscore in patients undergoing preoperative systemic therapy followed by surgery. The discovery and longitudinal validation cohorts together constituted the primary cohort. Patients within this cohort who underwent preoperative systemic therapy followed by surgery constituted the primary surgical cohort (n=143). This subset was used for clinicopathologic characterization and Cox analyses. For external validation, an independent temporal cohort of 147 patients with baseline serum samples was assembled between September 2024 and August 2025. Details of the patient selection process, exclusions, cohort overlap, and subgroup composition are shown in Supplementary Fig. 1.
Patients were classified into 3 categories based on metastatic burden: 1) Locally advanced gastric cancer (LAGC): patients with no distant metastases. 2) OMGC: defined according to criteria adapted from the AIO-FLOT3 trial [11]. OMGC was characterized by either (a) non-regional intra-abdominal lymph node metastasis, or any one of the following conditions (b), with or without (a): (b1) localized peritoneal metastasis: P0CY1, P1a, or P1b (peritoneal categories were defined according to the Japanese Classification of Gastric Carcinoma, 15th edition); (b2) liver metastasis: up to 5 metastatic lesions; (b3) lung metastasis: up to 5 unilateral metastatic lesions; (b4) ovarian metastasis; (b5) adrenal metastasis; (b6) single-region extra-abdominal lymph node metastasis; (b7) bone metastasis limited to a single radiation field. 3) Widely metastatic gastric cancer (WMGC): defined as the presence of metastatic disease beyond the criteria established for OMGC.
The study was approved by the Ethics Committee of Zhongshan Hospital, Fudan University (approval No. B2024-391). Written informed consent was obtained from all participants. The research adhered to the ethical principles outlined in the Declaration of Helsinki. The study was registered in the Chinese Clinical Trial Registry (ChiCTR2500096665; registered January 27, 2025).
Treatment strategy and clinical endpoints
In the primary surgical cohort, surgical decisions after systemic treatment were made by a multidisciplinary team. Staging laparoscopy and/or peritoneal cytology were performed when clinically indicated, particularly for suspected peritoneal disease. Patients were restaged by contrast-enhanced computed tomography or magnetic resonance imaging, and those with potentially resectable disease and adequate performance status underwent surgery. The type of gastrectomy and extent of lymphadenectomy were determined by tumor location and intraoperative findings, and local treatment was performed for metastatic lesions in selected patients. Baseline clinicopathologic characteristics, preoperative treatment, surgical procedures, and perioperative safety are summarized in Supplementary Table 1, and patterns of metastasis and local therapy for metastatic lesions among patients with metastatic GC are summarized in Supplementary Table 2.
Overall survival (OS) was defined as the interval from initial diagnosis (the date of pathological confirmation) to death from any cause. Patients alive at the last follow-up were censored on the date they were last known to be alive. Event-free survival (EFS) was defined as the interval from initial diagnosis to the first occurrence of disease progression, including local recurrence, new distant metastasis, progression of an existing metastatic or primary lesion, or death from any cause. Patients without an EFS event were censored at the date of the last disease assessment confirming event-free status. Major pathological response (MPR) was defined as <10% residual viable tumor (RVT) at the primary site, and pathological complete response was defined as the absence of RVT in both the primary tumor site and lymph nodes. Radiologic response was assessed according to RECIST version 1.1 using imaging performed within 14 days before surgery.
Serum sampling and proteomic profiling
Serum samples were prospectively collected after informed consent was obtained. Baseline serum samples were collected within 7 days before initiation of systemic therapy. Post-treatment serum samples were collected after completion of preoperative systemic therapy and within 7 days before surgery.
Proximity extension assays were performed using the Olink Target 96 Inflammation Panel. Protein levels were reported as normalized protein expression values on a log2 scale. Samples failing quality control were excluded. The OlinkAnalyze R package was used to perform gene set enrichment analysis (GSEA) of Gene Ontology (GO) gene sets.
TIPscore model construction
The TIPscore was developed using a multistep strategy for serum protein screening. The strategy incorporated ordinal logistic regression, extreme gradient boosting (XGBoost), and support vector machine with recursive feature elimination (SVM-RFE).
To model metastatic burden, Elastic Net regression was performed using the glmnet R package. The model was fitted with the alpha parameter set to 0.9 and tuned by 10-fold cross-validation using the lambda-minimum criterion. The rms R package was used to visualize model calibration.
The complete workflow for baseline serum protein screening, feature selection, and TIPscore construction is provided in the Supplementary Data 1 and Supplementary Tables 3, 4, 5.
Stage-membership probability modeling
Ordinal logistic regression was used to estimate the probability of belonging to each clinical state (LAGC, OMGC, or WMGC) as a continuous function of the TIPscore.
Conventional clinical biomarkers and composite indices
Conventional clinical biomarkers and composite indices were calculated using baseline laboratory data from the discovery cohort, collected at the same baseline time point as the serum samples used for TIPscore calculation. Serum tumor markers included carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9). The modified Glasgow Prognostic Score (mGPS) was calculated from serum C-reactive protein (CRP) and albumin levels. Patients with CRP ≤10 mg/L were assigned a score of 0; patients with CRP >10 mg/L were assigned a score of 1 when albumin was ≥35 g/L and a score of 2 when albumin was <35 g/L [17]. Additional composite indices included the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), monocyte-to-lymphocyte ratio (MLR), systemic immune-inflammation index (SII), systemic inflammation response index (SIRI), and prognostic nutritional index (PNI). These indices were calculated as follows: NLR = neutrophil count/lymphocyte count; PLR = platelet count/lymphocyte count; MLR = monocyte count/lymphocyte count; SII = platelet count × neutrophil count/lymphocyte count; SIRI = neutrophil count × monocyte count/lymphocyte count; and PNI = albumin (g/L) + 5 × lymphocyte count (109/L).
Tissue TIPscore
To validate the biological relevance of the TIPscore components in tumor tissue, we analyzed the TCGA-STAD cohort. A tissue TIPscore was derived using logistic regression to differentiate Stage II/III from Stage IV disease based on expression of the genes corresponding to the serum TIPscore components. This analysis was performed to evaluate the concordance between gene expression in tumor tissue and clinical metastatic burden.
RNA sequencing and analysis
Total RNA sequencing was performed on 20 frozen tumor specimens using the Illumina NovaSeq 6000 platform. These 20 primary tumor specimens were matched to corresponding serum TIPscore profiles and were obtained from patients in the primary cohort with available frozen tissue. The methods and dataset were previously described by our group [18]. Differentially expressed genes were identified using the EdgeR R package. Genes with an absolute log fold change greater than 2.5 and a P-value less than 0.05 were considered significantly differentially expressed. GSEA was conducted using the ClusterProfiler R package, focusing on GO-BP gene sets. Immune infiltration analysis was performed using the CIBERSORT-ABS method implemented in the Immunedeconv R package to estimate the relative abundance of immune cell types within the tumor microenvironment.
Statistical analysis
Two-group comparisons were performed using Student’s t-test for normally distributed variables and the Mann-Whitney U test for non-normally distributed variables. For comparisons involving more than 2 groups, one-way ANOVA was used for normally distributed variables. The Kruskal-Wallis test was used for non-normally distributed variables. Survival curves were generated using the Kaplan–Meier method, and the log-rank test was used to assess statistical significance; hazard ratios, 95% confidence intervals (CIs), and numbers at risk are reported for survival comparisons. The pROC R package (R Foundation for Statistical Computing, Vienna, Austria) was used to generate receiver operating characteristic curves and calculate the area under the curve (AUC) with a 95% CI. The multi-class AUC for classification across LAGC, OMGC, and WMGC was calculated using the Hand and Till method. Correlation analysis was performed using Spearman’s rank correlation.
Univariable and multivariable Cox proportional hazards models were constructed for OS and EFS in the primary surgical cohort. Variables with a univariable P-value <0.10 were entered into the multivariable models. To address potential immortal-time bias in analyses of the post-treatment TIPscore among patients undergoing surgery, landmark Cox analyses anchored at the date of surgery were performed.
For all analyses, a two-sided P-value <0.05 was considered statistically significant.
RESULTS
Construction of a biological continuum of metastatic disease
We reasoned that metastatic progression represents a continuous biological process shaped by tumor–host interactions [19,20], rather than discrete anatomic states. To capture this continuum, we used a discovery cohort of 107 baseline serum samples (LAGC=55, OMGC=22, WMGC=30) for high-dimensional profiling of 92 inflammation-related proteins using the Olink platform (Fig. 1, Supplementary Fig. 1, Table 1).
Fig. 1. Study design and conceptual framework. Patients with advanced GC were initially classified into LAGC, OMGC, and WMGC according to anatomic metastatic burden. The study included a discovery cohort with baseline serum samples (n=107), a temporal validation cohort with baseline serum samples (n=147), and a longitudinal validation cohort with post-treatment serum samples collected at re-evaluation after systemic therapy (n=104). Serum proteomic profiling was performed using the Olink Target 96 Inflammation Panel to develop the TIPscore, a biology-based model designed to move patient stratification from anatomy-based classification toward a biological continuum of metastatic disease. The TIPscore was subsequently evaluated for its ability to refine the biological stratification of OMGC, support prediction of clinical outcomes, and inform surgical decision-making. Fig. 1 was partially created with BioRender (Y.W. 2026. https://BioRender.com/rjp5ivi).
GC = gastric cancer; LAGC = locally advanced gastric cancer; OMGC = oligometastatic gastric cancer; WMGC = widely metastatic gastric cancer; TIPscore = Tumor-Induced Perturbation score; OS = overall survival; EFS = event-free survival.
Table 1. Clinicopathologic characteristics of patients with advanced gastric cancer across discovery and validation cohorts.
| Variables | Discovery cohort (n=107) | Longitudinal validation cohort (n=104) | Temporal validation cohort (n=147) | ||
|---|---|---|---|---|---|
| Age (yr), median [IQR] | 62.00 [54.50–68.00] | 62.00 [53.00–68.00] | 63.00 [57.50–70.50] | ||
| Gender | |||||
| Female | 30 (28.0) | 25 (24.0) | 35 (23.8) | ||
| Male | 77 (72.0) | 79 (76.0) | 112 (76.2) | ||
| Metastatic burden | |||||
| LAGC | 55 (51.4) | 71 (68.3) | 105 (71.4) | ||
| OMGC | 22 (20.6) | 30 (28.8) | 28 (19.0) | ||
| WMGC | 30 (28.0) | 3 (2.9) | 14 (9.5) | ||
| Organs with metastases | |||||
| 0 | 55 (51.4) | 71 (68.3) | 105 (71.4) | ||
| 1 | 23 (21.5) | 29 (27.9) | 22 (15.0) | ||
| 2 | 1 (0.9) | 1 (1.0) | 7 (4.8) | ||
| ≥3 | 28 (26.2) | 3 (2.9) | 13 (8.8) | ||
| Peritoneal metastases | |||||
| No | 68 (63.6) | 88 (84.6) | 122 (83.0) | ||
| Yes | 39 (36.4) | 16 (15.4) | 25 (17.0) | ||
| CY1 | 3 (2.8) | 2 (1.9) | 5 (3.4) | ||
| P1a | 3 (2.8) | 6 (5.8) | 5 (3.4) | ||
| P1b | 8 (7.5) | 5 (4.8) | 4 (2.7) | ||
| P1c | 25 (23.4) | 3 (2.9) | 11 (7.5) | ||
| Liver metastases | |||||
| No | 99 (92.5) | 99 (95.2) | 143 (97.3) | ||
| Yes | 8 (7.5) | 5 (4.8) | 4 (2.7) | ||
| Primary tumor location | |||||
| Stomach | 91 (85.0) | 80 (76.9) | 115 (78.2) | ||
| Gastroesophageal junction | 16 (15.0) | 24 (23.1) | 32 (21.8) | ||
| Borrmann classification | |||||
| Large type III/type IV | 23 (21.5) | 15 (14.4) | 28 (19.0) | ||
| Non-large type III/type IV | 84 (78.5) | 89 (85.6) | 119 (81.0) | ||
| Signet ring cell carcinoma | |||||
| No | 77 (72.0) | 85 (81.7) | 113 (76.9) | ||
| Yes | 28 (26.2) | 18 (17.3) | 32 (21.8) | ||
| Unknown | 2 (1.9) | 1 (1.0) | 2 (1.4) | ||
| Tumor differentiation | |||||
| G1/G2 | 26 (24.3) | 18 (17.3) | 34 (23.1) | ||
| G2–3/G3 | 80 (74.8) | 85 (81.7) | 110 (74.8) | ||
| Unknown | 1 (0.9) | 1 (1.0) | 3 (2.0) | ||
| Lauren classification | |||||
| Intestinal | 36 (33.6) | 39 (37.5) | 61 (41.5) | ||
| Diffuse | 35 (32.7) | 29 (27.9) | 32 (21.8) | ||
| Mixed | 26 (24.3) | 34 (32.7) | 43 (29.3) | ||
| Unknown | 10 (9.3) | 2 (1.9) | 11 (7.5) | ||
| HER2 overexpression | |||||
| No | 88 (82.2) | 91 (87.5) | 129 (87.8) | ||
| Yes | 9 (8.4) | 9 (8.7) | 13 (8.8) | ||
| Unknown | 10 (9.3) | 4 (3.8) | 5 (3.4) | ||
| MMR status | |||||
| dMMR | 5 (4.7) | 4 (3.8) | 16 (10.9) | ||
| pMMR | 70 (65.4) | 44 (42.3) | 126 (85.7) | ||
| Unknown | 32 (29.9) | 56 (53.8) | 5 (3.4) | ||
| PD-L1 expression | |||||
| CPS score <5 | 36 (33.6) | 31 (29.8) | 68 (46.3) | ||
| CPS score ≥5 | 38 (35.5) | 45 (43.3) | 64 (43.5) | ||
| Unknown | 33 (30.8) | 28 (26.9) | 15 (10.2) | ||
Values are presented as number (%) unless otherwise indicated.
Peritoneal metastases were classified according to the Japanese Classification of Gastric Carcinoma (15th edition): CY1, positive peritoneal cytology; P1a, metastasis limited to the stomach, omentum, or adjacent ligaments; P1b, metastasis to the upper abdominal parietal peritoneum or visceral peritoneum above the transverse colon; P1c, metastasis to the middle or lower abdominal peritoneum.
IQR = interquartile range; LAGC = locally advanced gastric cancer; OMGC = oligometastatic gastric cancer; WMGC = widely metastatic gastric cancer; HER2 = human epidermal growth factor receptor 2; MMR = mismatch repair; dMMR = deficient mismatch repair; pMMR = proficient mismatch repair; PD-L1 = programmed death-ligand 1; CPS = combined positive score.
Ordinal logistic regression using the ordered categories of LAGC–OMGC–WMGC identified 19 proteins significantly associated with progressive metastatic burden (Fig. 2A, Supplementary Table 3). To refine biomarker selection, XGBoost and SVM-RFE models were applied (Supplementary Table 4). The TIPscore was then constructed using Elastic Net regression and incorporated 3 core proteins: IL-22RA1, HGF, and 4E-BP1 (Fig. 2B, Supplementary Table 5).
Fig. 2. Development of the TIPscore and establishment of a biological continuum of metastatic disease. (A) Heatmap showing expression patterns of 19 candidate proteins identified by ordinal logistic regression across the clinical spectrum of LAGC, OMGC, and WMGC. Protein abundance was measured using the Olink Target 96 Inflammation Panel. (B) Feature selection and model construction using Elastic Net regression. Three proteins—IL-22RA1, HGF, and 4E-BP1—were incorporated into the TIPscore. (C) Distribution of baseline TIPscores across clinical groups in the discovery cohort, demonstrating a stepwise increase from LAGC to OMGC and WMGC. (D) Diagnostic performance of the baseline TIPscore in the discovery cohort for distinguishing LAGC from metastatic disease, identifying WMGC, and classifying patients across the entire metastatic spectrum. AUC values are shown. (E) Calibration plot comparing probabilities predicted by the Elastic Net model with observed clinical classifications. (F) Stage-membership probability modeling showing the predicted probability of belonging to the LAGC, OMGC, or WMGC state as a function of baseline TIPscore.
LAGC = locally advanced gastric cancer; OMGC = oligometastatic gastric cancer; WMGC = widely metastatic gastric cancer; TIPscore = Tumor-Induced Perturbation score; AUC = area under the curve.
Significance levels are indicated as follows: ***P<0.001, ****P<0.0001.
Baseline TIPscores showed a stepwise increase across LAGC, OMGC, and WMGC (Fig. 2C), positioning patients along a continuous biological spectrum of metastatic burden. The model achieved an AUC of 0.814 (95% CI, 0.723–0.894) for LAGC versus metastatic disease and 0.812 (95% CI, 0.714–0.894) for identifying WMGC, with good calibration (Fig. 2D and E). The multi-class AUC was 0.748 (95% CI, 0.677–0.812), indicating robust classification across the metastatic spectrum. The estimated probabilities of belonging to the LAGC, OMGC, and WMGC states varied progressively across the TIPscore continuum, with higher TIPscores corresponding to more advanced disease states (Fig. 2F). Together, these findings establish a biologically defined continuum of metastatic disease, providing a framework to redefine OMGC beyond conventional anatomic criteria.
Validation of the TIPscore as a robust and dynamic biological staging framework
To evaluate generalizability, the TIPscore was validated in an independent temporal cohort (n=147). Baseline TIPscores showed strong discriminatory performance across disease states (Fig. 3A), achieving an AUC of 0.711 (95% CI, 0.612–0.799) for distinguishing LAGC from metastatic disease, 0.810 (95% CI, 0.661–0.932) for identifying WMGC, and a multi-class AUC of 0.747 (95% CI, 0.639–0.842) (Fig. 3B). These results demonstrate that performance was consistent beyond the discovery cohort.
Fig. 3. Validation of the TIPscore across the metastatic spectrum and comparison with conventional indices. (A) Distribution of baseline TIPscores across LAGC, OMGC, and WMGC in the temporal validation cohort. (B) Diagnostic performance of the baseline TIPscore in the temporal validation cohort for classifying patients across the metastatic spectrum. (C) Distribution of post-treatment TIPscores across disease groups in the longitudinal validation cohort. (D) Multi-class AUC for ordinal LAGC–OMGC–WMGC classification comparing the TIPscore with CEA, CA19-9, NLR, PLR, MLR, SII, SIRI, PNI, and mGPS at baseline (discovery cohort) and post-treatment (longitudinal validation cohort). (E) Heatmap of standardized means for each marker across LAGC, OMGC, and WMGC.
TIPscore = Tumor-Induced Perturbation score; LAGC = locally advanced gastric cancer; OMGC = oligometastatic gastric cancer; WMGC = widely metastatic gastric cancer; SII = systemic immune-inflammation index; PLR = platelet-to-lymphocyte ratio; mGPS = modified Glasgow Prognostic Score; MLR = monocyte-to-lymphocyte ratio; SIRI = systemic inflammation response index; NLR = neutrophil-to-lymphocyte ratio; PNI = prognostic nutritional index; CA19-9 = carbohydrate antigen 19-9; CEA = carcinoembryonic antigen; ROC = receiver operating characteristic; AUC = area under the curve; SD = standard deviation.
Significance levels are indicated as follows: *P<0.05, **P<0.01, ****P<0.0001.
In a longitudinal validation cohort (n=104; LAGC=71, OMGC=30, WMGC=3), post-treatment TIPscores continued to differentiate disease states (Fig. 3C). Post-treatment samples in the longitudinal validation cohort showed lower TIPscores in OMGC than baseline samples in the discovery cohort (Supplementary Fig. 2). Although this was a cross-cohort rather than a within-patient paired comparison, the lower post-treatment TIPscores are consistent with a possible treatment-associated decrease. These findings support the biological staging performance of the TIPscore and suggest its potential utility for dynamic assessment of metastatic burden during treatment.
TIPscore complements anatomical criteria and conventional indices
To examine how the TIPscore relates to existing anatomic definitions, we evaluated baseline TIPscores within the OMGC group using combined data from the discovery and temporal validation cohorts. TIPscores did not differ significantly across major metastatic sites, including peritoneal, non-regional abdominal lymph node, distant lymph node, and liver metastases (P=0.862; Supplementary Fig. 3A). Among patients with peritoneal involvement, TIPscores increased across CY1, P1a, and P1b categories (P=0.035; Supplementary Fig. 3B), suggesting partial concordance with greater peritoneal disease extent. In contrast, TIPscores were not associated with OMEC-4 eligibility (P=0.777; Supplementary Fig. 3C) or CONVO-GC-1 category (P=0.163; Supplementary Fig. 3D). These findings indicate that the TIPscore is not fully explained by metastatic site or existing anatomy-based consensus definitions, supporting further evaluation of its prognostic value in subsequent analyses.
We next compared the TIPscore with routinely available serum tumor markers and systemic composite indices, including CEA, CA19-9, NLR, PLR, MLR, SII, SIRI, PNI, and mGPS. In both the discovery cohort at baseline and the longitudinal validation cohort after treatment, the TIPscore achieved the highest multi-class AUC for ordinal LAGC–OMGC–WMGC classification, exceeding all conventional comparators (Fig. 3D). In the standardized-mean analysis, the TIPscore also showed the most consistent monotonic increase across the metastatic spectrum, whereas conventional indices showed weaker or less consistent gradients (Fig. 3E).
Collectively, these results show that the TIPscore complements anatomic criteria and conventional clinical indices by providing a biologically informed characterization of metastatic burden.
Clinical utility of the TIPscore for prognostic and response stratification
We next evaluated the clinical utility of the TIPscore for outcome prediction as a biologically grounded stratification framework. At baseline, higher TIPscores were associated with worse survival in patients with LAGC (Fig. 4A), supporting its broader applicability for reflecting disease aggressiveness in GC.
Fig. 4. Clinical Utility of the TIPscore for Prognostic and Response Stratification. (A) Kaplan–Meier curves for OS and EFS in patients with LAGC and survival data stratified by median baseline TIPscore. (B) Upper panel: distribution of post-treatment TIPscores in the longitudinal validation cohort relative to the median cutoff. Lower panel: alignment of TIPscore status with survival outcome and clinical group. (C, D) Kaplan–Meier analysis of OS (C) and EFS (D) in the longitudinal validation cohort, including OMGC and LAGC subgroups, stratified by post-treatment TIPscore. (E) Receiver operating characteristic curves for predicting 1-year OS in the longitudinal validation cohort. Post-treatment TIPscore was compared with the conventional M0/M1 category in the overall cohort, and its discriminatory performance was further assessed within the OMGC subgroup. (F, G) Baseline (F) and post-treatment (G) TIPscores according to major pathological response (MPR) status in the discovery and longitudinal cohorts, respectively. (H) Distribution of baseline TIPscores according to RECIST 1.1 response categories in the combined OMGC and WMGC subgroup of the temporal validation cohort.
OS = overall survival; EFS = event-free survival; HR = hazard ratio; TIPscore = Tumor-Induced Perturbation score; LAGC = locally advanced gastric cancer; OMGC = oligometastatic gastric cancer; WMGC = widely metastatic gastric cancer; AUC = area under the curve; CI = confidence interval; MPR = major pathological response; PR = partial response; SD = stable disease; PD = progressive disease.
Post-treatment TIPscores demonstrated stronger prognostic value, with elevated scores associated with poorer long-term outcomes (Fig. 4B). Notably, in patients with OMGC undergoing curative-intent surgery, higher post-treatment TIPscores were associated with significantly worse OS (P=0.034) and EFS (P=0.015) (Fig. 4C and D), indicating that biological stratification can distinguish patients with OMGC who have different outcomes and may aid risk stratification among patients being considered for surgical intervention.
Importantly, the TIPscore showed superior performance in predicting 1-year OS in OMGC (AUC, 0.875; 95% CI, 0.717–1.000) and across all patients with GC (AUC, 0.850; 95% CI, 0.664–1.000), outperforming the binary M0/M1 category (AUC, 0.725; 95% CI, 0.507–0.943; Fig. 4E), further supporting its value in refining risk stratification beyond anatomic criteria.
We further assessed its ability to predict treatment response. Although baseline TIPscores showed limited association with pathological response (Fig. 4F), post-treatment TIPscores significantly predicted MPR in patients with OMGC (P<0.001), but not in LAGC (Fig. 4G). In the temporal validation cohort, baseline TIPscores also predicted radiologic response (RECIST 1.1) among patients with OMGC or WMGC. Lower scores were observed in patients with partial response or stable disease than in those with progressive disease (P=0.022; Fig. 4H).
To confirm the independent prognostic value of the TIPscore, we performed multivariable Cox regression adjusting for established prognostic factors in the primary surgical cohort (Supplementary Table 6). After adjustment, the post-treatment TIPscore remained independently associated with EFS, with a consistent trend for OS; metastatic burden and diffuse morphology were also independent predictors. Because survival was measured from diagnosis and patients had to survive to reach surgery, we conducted landmark Cox analyses anchored at the date of surgery to address immortal-time bias (Supplementary Table 7). In the surgery-based landmark models, a low post-treatment TIPscore remained associated with significantly better EFS and a consistent OS trend.
Collectively, these findings show that the TIPscore captures both survival outcomes and treatment responsiveness and, importantly, enables biologically informed stratification of patients with OMGC being considered for curative-intent surgery.
Clinical illustration of biological stratification in OMGC
To illustrate the clinical relevance of biological stratification, we examined 2 representative cases of OMGC with paired serum samples and longitudinal clinical data (Fig. 5).
Fig. 5. Clinical illustration of biologically distinct oligometastatic gastric cancer identified by the TIPscore. (A) Patient 66, who presented with positive peritoneal cytology (CY1), exhibited a low baseline TIPscore that remained stable during preoperative therapy and subsequently achieved a favorable pathological response with prolonged recurrence-free survival. (B) Patient 48, who presented with non-regional abdominal LN metastases, exhibited a high baseline TIPscore that further increased after therapy and subsequently developed rapid recurrence despite a pathological response at the primary tumor site similar to that of Patient 66.
TIPscore = Tumor-Induced Perturbation score; RVT = residual viable tumor; LN = lymph node.
Patient 66, a 38-year-old male presenting with positive peritoneal cytology (CY1), exhibited a low baseline TIPscore that remained stable following preoperative chemotherapy. Notably, the patient achieved a favorable pathological response (RVT=10%) and remained recurrence-free for over 2 years, with OS approaching 5 years (Fig. 5A).
In contrast, Patient 48, a 60-year-old woman with non-regional lymph node metastases, presented with a high baseline TIPscore that further increased after therapy. Despite achieving a tumor regression grade similar to that of Patient 66, she experienced rapid disease recurrence and died within 12 months of baseline (Fig. 5B).
Despite comparable pathological responses, these 2 patients demonstrated markedly different clinical outcomes, which were accurately reflected in their biological status as captured by the TIPscore. These findings are consistent with the view that biological stratification, rather than anatomic assessment alone, may help identify patients with more favorable biological profiles among those being considered for curative-intent surgery.
Biological basis of the TIPscore
To determine whether the TIPscore could noninvasively reflect the local tumor microenvironment, we analyzed transcriptomic profiles from tumor samples matched to corresponding serum TIPscore profiles and stratified by TIPscore levels. GSEA revealed that tumors from patients with low TIPscores were enriched for immune-related pathways (Supplementary Fig. 4A). Consistent with this finding, immune cell profiling demonstrated increased infiltration of CD8+ T cells, memory B cells, and plasma cells—immune subsets associated with favorable prognosis—in the low TIPscore group (Supplementary Fig. 4B). These findings indicate that the systemic TIPscore reflects the immune landscape of the local tumor microenvironment.
To further validate this biological association, we examined the expression patterns of the 3 TIPscore-related genes in the TCGA-STAD cohort. Using these components, we constructed a tissue-derived TIPscore by logistic regression. This tissue-based score effectively distinguished Stage II/III from Stage IV disease (Supplementary Table 8, Supplementary Fig. 5A and B) and was a significant prognostic indicator in advanced GC (Supplementary Fig. 5C).
Notably, its prognostic value was most evident in patients with Stage IIIC or IV disease—a subgroup in which conventional anatomic staging provides limited guidance for treatment decisions (Supplementary Fig. 5D), supporting the role of biology-driven stratification in advanced disease.
Among the TIPscore components, IL-22RA1 contributed most strongly to the model (Fig. 2B). Its tissue expression decreased in advanced stages, whereas its circulating signal contributed positively to the systemic TIPscore, indicating a divergence between tumor-tissue expression and the circulating signal. We observed a significant negative correlation between tissue IL22RA1 and its ligand IL22 specifically in Stage IV cases (Supplementary Fig. 5G), together with positive correlations of IL22RA1 with the proteases ADAM10 and ADAM17 (Supplementary Fig. 5H). These observations are hypothesis-generating and are discussed further below.
DISCUSSION
Conversion surgery offers a survival benefit for selected patients with metastatic GC [8]. Among these patients, those with OMGC are most likely to be considered for curative-intent surgery [11]. Therefore, improving risk stratification within OMGC is clinically important for identifying appropriate candidates for surgical intervention. However, the current anatomic definition of OMGC fails to capture the biological heterogeneity that determines surgical outcomes. To address this, we developed the TIPscore, a serum-based model that enables a biology-driven redefinition of OMGC by quantifying systemic tumor–host interactions, thereby supporting risk stratification of patients considered for surgery.
The definition of OMGC remains heterogeneous across trials, guidelines, and consensus statements [6,7,9,10,11,12,13,21], and this heterogeneity directly motivated our study. The most debated boundaries involve liver and peritoneal disease. For liver metastases, OMEC-4 restricts OMGC to 3 or fewer lesions [7], whereas the RENAISSANCE trial permits up to 5 lesions [13]. For peritoneal disease, OMEC-4 excludes peritoneal metastasis [7], whereas Japanese guidelines and clinical practice have considered gastrectomy after chemotherapy for selected patients with CY1 or limited peritoneal disease [8,10,22]. We therefore adopted a broad OMGC definition to include most patients with potentially oligometastatic disease. This approach was not intended to imply that all such patients are equally suitable for surgery, but rather to establish a broad anatomic reference population in which biological heterogeneity could be evaluated. Within this cohort, TIPscores were not fully explained by metastatic site, OMEC-4 eligibility, or CONVO-GC-1 category [8,22], while showing partial concordance with greater peritoneal disease extent. Together with its prognostic value in subsequent analyses, these findings suggest that anatomic criteria alone provide an incomplete representation of the underlying metastatic state and support the need for biology-informed stratification.
This framework is grounded in the established role of chronic inflammation and systemic tumor–host interactions in cancer progression [23]. Metastasis is not only a property of tumor cells but also reflects reciprocal interactions between the tumor, immune system, and systemic inflammatory state [24,25,26,27]. By quantifying these systemic interactions, the TIPscore provides biological insight at the decision point between systemic therapy and surgical intervention. Unlike conventional imaging, which often fails to capture the biological complexity of metastatic disease [28,29], the post-treatment TIPscore showed strong performance in predicting 1-year OS in OMGC. These findings suggest that systemic biological status after therapy may complement anatomic reassessment in post-treatment risk stratification.
A key conceptual advance of this study is the redefinition of OMGC as a biologically dynamic state rather than a fixed anatomic category. Stage-membership probability modeling showed that OMGC occupied an intermediate position along the metastatic continuum. The observed reduction in TIPscore, particularly in OMGC, suggests that effective therapy may partially reverse systemic tumor-induced perturbation and shift patients toward a less advanced biological state. This dynamic behavior distinguishes the TIPscore from static anatomic definitions and supports its potential use as a real-time marker of treatment-induced biological control.
This dynamic biological information may help address an important clinical dilemma in OMGC: radiologic or pathological response does not always translate into durable survival. In our cohort, lower post-treatment TIPscores were associated with better OS and EFS among patients with OMGC undergoing curative-intent surgery, whereas higher post-treatment TIPscores identified patients with poorer outcomes despite surgical treatment. Multivariable and surgery-based landmark Cox analyses further supported the independent prognostic value of the post-treatment TIPscore, particularly for EFS. These findings suggest that the TIPscore may help risk-stratify patients with OMGC being considered for curative-intent surgery.
Our findings also align with the broader clinical movement toward response-adapted conversion strategies. For example, JCOG2301 selects candidates for conversion surgery based on treatment response [30]. Although such approaches rely on observable clinical response, the TIPscore provides a biology-based metric that captures the underlying tumor–host state. The case illustrations further support this point: 2 patients with apparently comparable local pathological regression had markedly different systemic TIPscore trajectories and clinical outcomes. These observations highlight that local tumor regression and systemic biological control are related but not interchangeable, and that integrating serum-based biological assessment may improve risk stratification after systemic therapy.
The performance difference between the discovery and temporal validation cohorts also warrants consideration. The AUC for distinguishing LAGC from metastatic disease was lower in the temporal validation cohort than in the discovery cohort. This likely reflects differences in cohort composition and disease spectrum. Compared with the discovery cohort, the temporal validation cohort included a larger proportion of LAGC patients, while its metastatic subgroup contained relatively fewer WMGC cases and a higher proportion of OMGC cases. As a result, the binary comparison between LAGC and metastatic disease was more strongly driven by the biologically intermediate OMGC subgroup, rather than by clearly disseminated WMGC. This may have increased overlap between clinically defined LAGC and metastatic disease, thereby reducing the apparent binary discrimination. Importantly, the ordinal discrimination across LAGC, OMGC, and WMGC, as well as the ability to identify WMGC, was preserved in the temporal validation cohort, supporting the robustness of the TIPscore as a continuum-based biological staging marker.
The association between the TIPscore and tumor biology further supports its biological relevance. Tumors from patients with low TIPscores showed enrichment of immune-related pathways and higher infiltration of immune cell populations associated with favorable prognosis, suggesting that systemic tumor–host perturbation reflected in serum may, at least in part, mirror the tumor immune microenvironment. The tissue-derived TIPscore in TCGA-STAD further distinguished Stage II/III from Stage IV disease and showed prognostic relevance in advanced-stage GC, supporting cross-platform biological consistency between serum protein signals and tumor-tissue gene expression.
Among the 3 TIPscore components, IL-22RA1 made the strongest contribution to the model. Interestingly, tissue IL-22RA1 expression decreased in advanced-stage tumors, whereas circulating IL-22RA1 contributed positively to the systemic TIPscore. This divergence suggests that serum IL-22RA1 may not simply mirror tumor-tissue transcriptional expression, but may instead reflect more complex regulation of the IL-22 signaling axis within the metastatic tumor–host environment. IL-22 signaling has been implicated in tumor progression and metastatic colonization [31,32], and receptor ectodomain shedding is a recognized mechanism by which circulating receptor levels may diverge from tissue expression [33,34]. In this context, the negative correlation between tissue IL22RA1 and IL22 in Stage IV disease, together with the positive associations of IL22RA1 with ADAM10 and ADAM17, raises the possibility that circulating IL-22RA1 may reflect protease-related remodeling or receptor shedding within the metastatic microenvironment. However, this interpretation remains hypothesis-generating because it has not been experimentally validated in the present study. Further functional studies are needed to determine whether IL-22RA1 actively contributes to metastatic progression or primarily marks systemic tumor–host perturbation.
Several limitations warrant consideration. First, the TIPscore was derived from an inflammation-focused panel and may not capture the full spectrum of tumor–host interactions. Broader proteomic, genomic, and cellular immune profiling may further refine this framework. Second, although the study included discovery, longitudinal validation, and temporal validation cohorts, OMGC subgroup sizes were modest, the longitudinal validation cohort contained only 3 WMGC patients, and the number of paired longitudinal serum samples was limited. Small-subgroup comparisons should therefore be interpreted cautiously. Third, the temporal validation cohort was derived from the same institution, and multicenter external validation is required. Finally, because no contemporaneous nonsurgical OMGC comparator was available, the present data support biological risk stratification among surgical candidates rather than a demonstrated predictive benefit of surgery. Future prospective, multicenter studies are needed to confirm these findings and refine the TIPscore before clinical use.
In conclusion, redefining OMGC from an anatomic to a biological entity represents a critical step toward precision surgical oncology. By quantifying systemic tumor–host perturbation, the TIPscore provides a real-time liquid biopsy framework for informing treatment decisions in advanced GC. This approach may support risk stratification of biologically appropriate candidates being considered for curative-intent surgery, pending prospective validation.
Footnotes
Funding: This study was supported by the Noncommunicable Chronic Diseases–National Science and Technology Major Project (2023ZD0510600, to G.Y.), the Science and Technology Commission of Shanghai Municipality (23ZR1410700 and 23Y11902100, to T.Z.), the Natural Science Foundation of Fujian Province (2023J011692, to T.Z.), the National Natural Science Foundation of China (82273303, to S.Y.; 82102503, to W.Z.), and the Special Project for Clinical Research of Zhongshan Hospital (ZSLCYJ202335, to T.Z.).
Conflict of Interest: No potential conflict of interest relevant to this article was reported.
Data Availability Statement: The raw serum proteomics data generated in this study have been deposited in the National Genomics Data Center (NGDC; https://ngdc.cncb.ac.cn) under accession numbers OMIX001799, OMIX001800, and OMIX008466, and are publicly available. Tumor RNA-sequencing data were obtained from the TCGA-STAD dataset through the Genomic Data Commons portal (https://portal.gdc.cancer.gov/). Additional information is available from the corresponding author upon reasonable request.
Meeting Presentation: A preliminary summary of these findings was presented at the AACR Annual Meeting 2026 in San Diego, CA, USA (Abstract ID: 3769).
- Conceptualization: T.Z., W.X.
- Data curation: Z.H., C.Y., L.T.
- Formal analysis: W.Y.
- Funding acquisition: W.Z., G.Y., S.Y., T.Z.
- Investigation: W.Z., F.M.
- Methodology: G.Y.
- Project administration: F.M., S.Z., S.K.
- Resources: W.Z., C.Y., L.T.
- Supervision: S.Y., W.X.
- Validation: G.Y.
- Visualization: W.Y.
- Writing - original draft: W.Y.
- Writing - review & editing: W.Y., T.Z.
SUPPLEMENTARY MATERIALS
Supplementary Materials
Baseline clinicopathologic characteristics, treatment, and surgical safety of the primary surgical cohort
Pattern of metastasis and local therapy to metastatic lesions among metastatic patients in the primary surgical cohort
Ordinal logistic regression analysis of baseline serum proteins associated with metastatic burden (related to Fig. 2A)
Serum protein screening using XGBoost and SVM-RFE
Development of the serum TIPscore using an Elastic Net regression model (related to Fig. 2B)
Univariable and multivariable Cox regression analyses for event-free and overall survival in the primary surgical cohort
Landmark (time-dependent) Cox sensitivity analyses of post-treatment TIPscore for survival in the longitudinal validation cohort
Development of the tissue-derived (tumor) TIPscore using a logistic regression model (related to Supplementary Fig. 5A)
Study flow diagram of the patient cohorts.
Cross-cohort comparison of baseline and post-treatment TIPscore distributions.
Baseline TIPscore across anatomical metastatic subgroups and existing OMGC classification systems.
Characterization of the tumor microenvironment according to post-treatment TIPscore.
Validation and biological exploration of the tissue-derived TIPscore in the TCGA-STAD cohort.
References
- 1.Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74:229–263. doi: 10.3322/caac.21834. [DOI] [PubMed] [Google Scholar]
- 2.Zeng H, Ran X, An L, Zheng R, Zhang S, Ji JS, et al. Disparities in stage at diagnosis for five common cancers in China: a multicentre, hospital-based, observational study. Lancet Public Health. 2021;6:e877–e887. doi: 10.1016/S2468-2667(21)00157-2. [DOI] [PubMed] [Google Scholar]
- 3.Hellman S, Weichselbaum RR. Oligometastases. J Clin Oncol. 1995;13:8–10. doi: 10.1200/JCO.1995.13.1.8. [DOI] [PubMed] [Google Scholar]
- 4.Ajani JA, D’Amico TA, Bentrem DJ, Corvera CU, Das P, Enzinger PC, et al. Gastric cancer, version 2.2025, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw. 2025;23:169–191. doi: 10.6004/jnccn.2025.0022. [DOI] [PubMed] [Google Scholar]
- 5.Narita Y, Muro K, Takahari D. Practical management of oligometastatic gastric cancer. ESMO Gastrointest Oncol. 2024;6:100108. doi: 10.1016/j.esmogo.2024.100108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kroese TE, van Laarhoven HWM, Schoppman SF, Deseyne PRAJ, van Cutsem E, Haustermans K, et al. Definition, diagnosis and treatment of oligometastatic oesophagogastric cancer: a Delphi consensus study in Europe. Eur J Cancer. 2023;185:28–39. doi: 10.1016/j.ejca.2023.02.015. [DOI] [PubMed] [Google Scholar]
- 7.Kroese TE, Bronzwaer S, van Rossum PSN, Schoppman SF, Deseyne PRAJ, van Cutsem E, et al. European clinical practice guidelines for the definition, diagnosis, and treatment of oligometastatic esophagogastric cancer (OMEC-4) Eur J Cancer. 2024;204:114062. doi: 10.1016/j.ejca.2024.114062. [DOI] [PubMed] [Google Scholar]
- 8.Yasufuku I, Kinoshita T, Bando E, Kurokawa Y, Fujisaki M, Ojima H, et al. Correction: Safety, survival outcomes, and prognostic factors in conversion surgery for advanced gastric cancer: a sub-analysis of Japanese patients in the CONVO-GC-1 study. Gastric Cancer. 2026;29:191. doi: 10.1007/s10120-025-01709-4. [DOI] [PubMed] [Google Scholar]
- 9.Wang FH, Zhang XT, Tang L, Wu Q, Cai MY, Li YF, et al. The Chinese Society of Clinical Oncology (CSCO): Clinical guidelines for the diagnosis and treatment of gastric cancer, 2023. Cancer Commun (Lond) 2024;44:127–172. doi: 10.1002/cac2.12516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Japanese Gastric Cancer Association. Japanese gastric cancer treatment guidelines 2025 (7th edition) Gastric Cancer. 2026;29:271–299. doi: 10.1007/s10120-025-01698-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Al-Batran SE, Homann N, Pauligk C, Illerhaus G, Martens UM, Stoehlmacher J, et al. Effect of neoadjuvant chemotherapy followed by surgical resection on survival in patients with limited metastatic gastric or gastroesophageal junction cancer: the AIO-FLOT3 trial. JAMA Oncol. 2017;3:1237–1244. doi: 10.1001/jamaoncol.2017.0515. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Cui Y, Yu Y, Zheng S, Ying J, Du Y, Wang Y, et al. Does resection after neoadjuvant chemotherapy of docetaxel, oxaliplatin, and S-1 (DOS regimen) benefit for gastric cancer patients with single non-curable factor? a multicenter, prospective cohort study (Neo-REGATTA) BMC Cancer. 2023;23:308. doi: 10.1186/s12885-023-10773-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Al-Batran SE, Goetze TO, Mueller DW, Vogel A, Winkler M, Lorenzen S, et al. The RENAISSANCE (AIO-FLOT5) trial: effect of chemotherapy alone vs. chemotherapy followed by surgical resection on survival and quality of life in patients with limited-metastatic adenocarcinoma of the stomach or esophagogastric junction - a phase III trial of the German AIO/CAO-V/CAOGI. BMC Cancer. 2017;17:893. doi: 10.1186/s12885-017-3918-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Al-Batran SE, Lorenzen S, Riera J, Caca K, Mueller C, Stange DE, et al. Effect of chemotherapy/targeted therapy alone vs. chemotherapy/targeted therapy followed by radical surgical resection on survival and quality of life in patients with limited-metastatic adenocarcinoma of the stomach or esophagogastric junction: The IKF-575/RENAISSANCE phase III trial. J Clin Oncol. 2024;42(17 Suppl):LBA4001 [Google Scholar]
- 15.Morgagni P, Bencivenga M, Carneiro F, Cascinu S, Derks S, Di Bartolomeo M, et al. International consensus on the management of metastatic gastric cancer: step by step in the foggy landscape : Bertinoro Workshop, November 2022. Gastric Cancer. 2024;27:649–671. doi: 10.1007/s10120-024-01479-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hellman S. Karnofsky memorial lecture. Natural history of small breast cancers. J Clin Oncol. 1994;12:2229–2234. doi: 10.1200/JCO.1994.12.10.2229. [DOI] [PubMed] [Google Scholar]
- 17.McMillan DC, Crozier JE, Canna K, Angerson WJ, McArdle CS. Evaluation of an inflammation-based prognostic score (GPS) in patients undergoing resection for colon and rectal cancer. Int J Colorectal Dis. 2007;22:881–886. doi: 10.1007/s00384-006-0259-6. [DOI] [PubMed] [Google Scholar]
- 18.Tang Z, Gu Y, Shi Z, Min L, Zhang Z, Zhou P, et al. Multiplex immune profiling reveals the role of serum immune proteomics in predicting response to preoperative chemotherapy of gastric cancer. Cell Rep Med. 2023;4:100931. doi: 10.1016/j.xcrm.2023.100931. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lambert AW, Pattabiraman DR, Weinberg RA. Emerging biological principles of metastasis. Cell. 2017;168:670–691. doi: 10.1016/j.cell.2016.11.037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Katipally RR, Pitroda SP, Juloori A, Chmura SJ, Weichselbaum RR. The oligometastatic spectrum in the era of improved detection and modern systemic therapy. Nat Rev Clin Oncol. 2022;19:585–599. doi: 10.1038/s41571-022-00655-9. [DOI] [PubMed] [Google Scholar]
- 21.Fujitani K, Yang HK, Mizusawa J, Kim YW, Terashima M, Han SU, et al. Gastrectomy plus chemotherapy versus chemotherapy alone for advanced gastric cancer with a single non-curable factor (REGATTA): a phase 3, randomised controlled trial. Lancet Oncol. 2016;17:309–318. doi: 10.1016/S1470-2045(15)00553-7. [DOI] [PubMed] [Google Scholar]
- 22.Yoshida K, Yasufuku I, Terashima M, Young Rha S, Moon Bae J, Li G, et al. International retrospective cohort study of conversion therapy for stage IV gastric cancer 1 (CONVO-GC-1) Ann Gastroenterol Surg. 2021;6:227–240. doi: 10.1002/ags3.12515. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hanahan D. Hallmarks of cancer: new dimensions. Cancer Discov. 2022;12:31–46. doi: 10.1158/2159-8290.CD-21-1059. [DOI] [PubMed] [Google Scholar]
- 24.Dvorak HF. Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing. N Engl J Med. 1986;315:1650–1659. doi: 10.1056/NEJM198612253152606. [DOI] [PubMed] [Google Scholar]
- 25.Grivennikov SI, Greten FR, Karin M. Immunity, inflammation, and cancer. Cell. 2010;140:883–899. doi: 10.1016/j.cell.2010.01.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Deyell M, Garris CS, Laughney AM. Cancer metastasis as a non-healing wound. Br J Cancer. 2021;124:1491–1502. doi: 10.1038/s41416-021-01309-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Martin P, Pardo-Pastor C, Jenkins RG, Rosenblatt J. Imperfect wound healing sets the stage for chronic diseases. Science. 2024;386:eadp2974. doi: 10.1126/science.adp2974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Schmidt T, Sicic L, Blank S, Becker K, Weichert W, Bruckner T, et al. Prognostic value of histopathological regression in 850 neoadjuvantly treated oesophagogastric adenocarcinomas. Br J Cancer. 2014;110:1712–1720. doi: 10.1038/bjc.2014.94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sun C, Niu P, Zhang X, Zhao L, Wang W, Luan X, et al. Concurrent clinical and pathological response predicts favorable prognosis of patients with gastric cancer after neoadjuvant therapy: a real-world study. BMC Cancer. 2023;23:996. doi: 10.1186/s12885-023-11508-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Japan Clinical Oncology Group. A randomized controlled trial comparing conversion surgery with palliative chemotherapy in patients with initially unresectable cStage IVB/pStage IV advanced gastric cancer who presented remarkable response to chemotherapy (JCOG2301) [Internet] Tokyo: Japan Clinical Oncology Group; 2024. [cited 2024 Dec 15]. Available from: https://jcog.jp/document/2301.pdf. [Google Scholar]
- 31.Giannou AD, Kempski J, Shiri AM, Lücke J, Zhang T, Zhao L, et al. Tissue resident iNKT17 cells facilitate cancer cell extravasation in liver metastasis via interleukin-22. Immunity. 2023;56:125–142.e12. doi: 10.1016/j.immuni.2022.12.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Hunzeker ZE, Zhao L, Kim AM, Parker JM, Zhu Z, Xiao H, et al. The role of IL-22 in cancer. Med Oncol. 2024;41:240. doi: 10.1007/s12032-024-02481-8. [DOI] [PubMed] [Google Scholar]
- 33.Miller MA, Sullivan RJ, Lauffenburger DA. Molecular pathways: receptor ectodomain shedding in treatment, resistance, and monitoring of cancer. Clin Cancer Res. 2017;23:623–629. doi: 10.1158/1078-0432.CCR-16-0869. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Park EJ, Lee CW. Soluble receptors in cancer: mechanisms, clinical significance, and therapeutic strategies. Exp Mol Med. 2024;56:100–109. doi: 10.1038/s12276-023-01150-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Materials
Baseline clinicopathologic characteristics, treatment, and surgical safety of the primary surgical cohort
Pattern of metastasis and local therapy to metastatic lesions among metastatic patients in the primary surgical cohort
Ordinal logistic regression analysis of baseline serum proteins associated with metastatic burden (related to Fig. 2A)
Serum protein screening using XGBoost and SVM-RFE
Development of the serum TIPscore using an Elastic Net regression model (related to Fig. 2B)
Univariable and multivariable Cox regression analyses for event-free and overall survival in the primary surgical cohort
Landmark (time-dependent) Cox sensitivity analyses of post-treatment TIPscore for survival in the longitudinal validation cohort
Development of the tissue-derived (tumor) TIPscore using a logistic regression model (related to Supplementary Fig. 5A)
Study flow diagram of the patient cohorts.
Cross-cohort comparison of baseline and post-treatment TIPscore distributions.
Baseline TIPscore across anatomical metastatic subgroups and existing OMGC classification systems.
Characterization of the tumor microenvironment according to post-treatment TIPscore.
Validation and biological exploration of the tissue-derived TIPscore in the TCGA-STAD cohort.