Jiang SW, Zhang CG, Wang KD, Shang KP, Yu PJ, Wang HL. Chemotherapy regimens with immunotherapy in advanced gastric cancer: A prognostic study using Cox modeling. World J Gastrointest Oncol 2026; 18(9): 121765 [DOI: 10.4251/wjgo.121765]
Corresponding Author of This Article
Huan-Li Wang, Professor, Qinghai University Affiliated Hospital Tumor Day Chemotherapy Center, Qinghai University Affiliated Hospital, No. 29 Tongren Road, Chengxi District, Xining 810000, Qinghai Province, China. 229572196@qq.com
Research Domain of This Article
Oncology
Article-Type of This Article
research-article
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA
Share the Article
Jiang SW, Zhang CG, Wang KD, Shang KP, Yu PJ, Wang HL. Chemotherapy regimens with immunotherapy in advanced gastric cancer: A prognostic study using Cox modeling. World J Gastrointest Oncol 2026; 18(9): 121765 [DOI: 10.4251/wjgo.121765]
Shao-Wei Jiang, Department of Radiation Oncology, Qinghai University Affiliated Hospital, Xining 810000, Qinghai Province, China
Chen-Guang Zhang, Department of Otolaryngology, Qinghai University Affiliated Hospital, Xining 810000, Qinghai Province, China
Ke-Di Wang, Kun-Peng Shang, Department of Gastrointestinal Oncology, Qinghai University Affiliated Hospital, Qinghai University, School of Clinical Medicine, Xining 810000, Qinghai Province, China
Peng-Jie Yu, Department of Gastrointestinal Oncology, Qinghai University Affiliated Hospital, Xining 810000, Qinghai Province, China
Huan-Li Wang, Qinghai University Affiliated Hospital Tumor Day Chemotherapy Center, Qinghai University Affiliated Hospital, Xining 810000, Qinghai Province, China
Co-first authors: Shao-Wei Jiang and Chen-Guang Zhang.
Co-corresponding authors: Peng-Jie Yu and Huan-Li Wang.
Author contributions: Jiang SW and Zhang CG were responsible for conceptualization, investigation, methodology, software, formal analysis, project administration, resources, data curation, writing original draft as co-first authors; Wang KD was responsible for software, writing original draft; Shang KP was responsible for data curation, writing original draft; Yu PJ and Wang HL were responsible for writing review editing, visualization, validation, funding acquisition, supervision as co-corresponding authors.
Supported by Qinghai Provincial Science and Technology Program, No. 2023-ZJ-787.
Institutional review board statement: This study complied with the Declaration of Helsinki and was approved by the Ethics Committee of Qinghai University Affiliated Hospital (No. SL-2022-035).
Informed consent statement: All participants provided informed consent.
Conflict-of-interest statement: All authors declare no conflict of interest in publishing the manuscript.
Data sharing statement: The data supporting the findings of this study are available from the corresponding author upon reasonable request, subject to institutional and ethical regulations.
Corresponding author: Huan-Li Wang, Professor, Qinghai University Affiliated Hospital Tumor Day Chemotherapy Center, Qinghai University Affiliated Hospital, No. 29 Tongren Road, Chengxi District, Xining 810000, Qinghai Province, China. 229572196@qq.com
Received: April 1, 2026 Revised: April 20, 2026 Accepted: June 2, 2026 Published online: September 15, 2026 Processing time: 161 Days and 23.2 Hours
Abstract
BACKGROUND
Locally advanced gastric cancer (GC) remains associated with a substantial risk of recurrence and poor long-term survival despite advances in multimodal treatment. Perioperative chemotherapy is a standard strategy, but the integration of immunotherapy into neoadjuvant treatment has shown promising potential. However, real-world comparative evidence regarding the efficacy and safety of different perioperative regimens remains limited.
AIM
To compare the efficacy and safety of three neoadjuvant treatment regimens – nab-paclitaxel plus oxaliplatin and S-1, oxaliplatin plus leucovorin and fluorouracil, and S-1 combined with sintilimab and oxaliplatin – in patients with locally advanced GC. Additionally, independent prognostic factors associated with progression-free survival (PFS) were identified, and a predictive model was developed to enable individualized risk stratification and prognostic assessment.
METHODS
This retrospective study included 298 patients with locally advanced GC who met the inclusion and exclusion criteria. Patients were randomly divided into a training set and a validation set at a 7:3 ratio using a fixed random seed. In the training set, least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation was used to select variables on the basis of the λ.1se criterion. Variables with nonzero coefficients were entered into a multivariable Cox proportional hazards model to identify independent factors associated with PFS, with hazard ratios (HRs) and 95%CIs calculated. The model was developed in the training set and validated in the validation set. Short-term efficacy, survival outcomes, and adverse events were compared among the three groups. Model performance was evaluated using receiver operating characteristic curves and calibration plots.
RESULTS
LASSO regression identified five variables with nonzero coefficients, including tumor differentiation, N stage, TNM stage, Response Evaluation Criteria in Solid Tumors 1.1 response, and tumor regression grade. Among these, TNM stage IIIC had the largest coefficient, indicating that it had the strongest impact on prognosis. These variables were subsequently included in a multivariable Cox proportional hazards model. The results demonstrated that poor differentiation (HR = 1.86; 95%CI: 1.19-2.91; P = 0.006), lymph node metastasis (HR = 1.69; 95%CI: 1.11-2.57; P = 0.013), and locally advanced clinical stage (cTNM stage IIIC; HR = 3.94; 95%CI: 2.47-6.28; P < 0.001) were independent risk factors for PFS in patients with GC. In contrast, a favorable response based on Response Evaluation Criteria in Solid Tumors 1.1 (HR = 0.65; 95%CI: 0.46-0.92; P = 0.016) and a lower tumor regression grade (HR = 0.56; 95%CI: 0.39-0.82; P = 0.003) were identified as protective factors.
CONCLUSION
This study demonstrated that compared with plus oxaliplatin and S-1 and oxaliplatin plus leucovorin and fluorouracil, the S-1 combined with sintilimab and oxaliplatin regimen resulted in a greater pathological response rate in patients with locally advanced GC and resulted in superior outcomes in terms of both PFS and overall survival, with an overall acceptable safety profile. The predictive model constructed based on LASSO and multivariable Cox regression exhibited good discrimination and calibration and may serve as a useful tool for post-treatment prognostic assessment and individualized risk stratification.
Core Tip: This study compares three commonly used neoadjuvant regimens (S-1 combined with oxaliplatin and sintilimab, plus oxaliplatin and S-1, and oxaliplatin plus leucovorin and fluorouracil) in advanced gastric cancer and demonstrates that the addition of a programmed cell death protein 1 inhibitor improves pathological response without increasing unacceptable toxicity. Furthermore, a prognostic model integrating treatment response (Response Evaluation Criteria in Solid Tumors 1.1 and tumor regression grade) with clinicopathological factors provides reliable prediction of progression-free survival, highlighting the value of response-based variables in individualized risk stratification.
Citation: Jiang SW, Zhang CG, Wang KD, Shang KP, Yu PJ, Wang HL. Chemotherapy regimens with immunotherapy in advanced gastric cancer: A prognostic study using Cox modeling. World J Gastrointest Oncol 2026; 18(9): 121765
Gastric cancer (GC) is one of the most common malignancies of the digestive system worldwide, with high incidence and mortality rates[1]. Despite advances in surgical techniques and perioperative multimodal therapies in recent years, the prognosis of patients with advanced GC remains unsatisfactory, with a high risk of recurrence and metastasis[2,3]. Therefore, optimizing neoadjuvant treatment strategies and achieving precise prognostic assessment have become key focuses of current clinical research[4]. At present, chemotherapy regimens based on oxaliplatin combined with fluoropyrimidines [such as S-1 combined with oxaliplatin and sintilimab (SOX + XDL) and FOLFOX (oxaliplatin plus leucovorin and fluorouracil)] have been widely used in the neoadjuvant treatment of advanced GC[5]. With the development of immunotherapy, programmed cell death protein 1 (PD-1) inhibitors combined with chemotherapy have emerged as a promising therapeutic approach, with several studies demonstrating improved tumor response rates and survival outcomes[6]. In addition, nab-paclitaxel-based combination regimens have shown favorable antitumor activity in certain patient populations[7]. However, direct comparative studies evaluating the efficacy and safety of different neoadjuvant regimens remain limited, and clinical decision-making still lacks high-quality evidence. On the other hand, traditional prognostic evaluation mainly relies on clinical staging and pathological characteristics, which may not fully capture individual heterogeneity among patients[8]. In recent years, statistical modeling-based prediction models have been increasingly applied in oncological prognostic assessment. The least absolute shrinkage and selection operator (LASSO) regression can effectively identify key variables in high-dimensional data, and its integration with Cox proportional hazards models facilitates the development of robust survival prediction models[9,10]. Nevertheless, for patients with advanced GC undergoing perioperative treatment, predictive models that incorporate both treatment response indicators [Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 and tumor regression grade (TRG)] and conventional clinical features remain to be further explored[11]. Based on this background, the present study aimed to compare the efficacy and safety of three neoadjuvant regimens – nab-paclitaxel plus oxaliplatin and S-1 (PSOX), FOLFOX, and SOX + XDL – in patients with locally advanced GC. Furthermore, key prognostic factors were identified using LASSO regression and multivariable Cox analysis, and a progression-free survival (PFS) prediction model was developed to provide evidence for individualized treatment decision-making.
MATERIALS AND METHODS
Patient information
A total of 298 patients with locally advanced GC treated at the Department of Gastrointestinal Oncology, Qinghai University Affiliated Hospital, between March 2018 and March 2023 were retrospectively enrolled. Among them, 106 patients received SOX + XDL, 150 received PSOX, and 42 received FOLFOX. All patients were randomly divided into a training set (70%) and a validation set (30%) using the sample function in R. The training set was used for model development, while the validation set was used for internal validation. All patients received 2-4 cycles of neoadjuvant chemotherapy and 2-4 cycles of adjuvant chemotherapy, for a total of 6-8 cycles, combined with D2 radical gastrectomy. Patients were followed up regularly for 3 years postoperatively or until the occurrence of predefined endpoints. The treatment plan can be found in Supplementary material.
Inclusion and exclusion criteria
All patients met the following inclusion criteria: (1) Histologically confirmed primary gastric adenocarcinoma diagnosed by imaging and endoscopic biopsy, staged as IIB-IIIC according to the 8th edition of the American Joint Committee on Cancer TNM staging system of the Union for International Cancer Control, without other malignancies or distant metastasis, and achieving R0 resection (no residual tumor macroscopically or microscopically); (2) Receipt of 2-4 cycles of neoadjuvant chemotherapy and 2-4 cycles of adjuvant chemotherapy (total 6-8 cycles), with all surgical procedures performed at our institution in accordance with National Comprehensive Cancer Network and Chinese Society of Clinical Oncology guidelines; measurable primary tumor lesions on computed tomography or magnetic resonance imaging with postoperative pathological confirmation; and (3) An Eastern Cooperative Oncology Group performance status ≤ 1, with adequate hepatic, renal, hematologic, and cardiopulmonary function to tolerate chemotherapy. Patients were excluded if they had allergies to chemotherapy agents or contraindications to chemotherapy, severe comorbid conditions (such as infectious diseases, gastrointestinal bleeding, pyloric obstruction, or gastrointestinal perforation), a history of other malignancies treated with radiotherapy, chemotherapy, biological therapy, or surgery, or incomplete clinical or imaging data that precluded accurate tumor measurement. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Qinghai University Affiliated Hospital, and written informed consent was obtained from all participants.
Evaluation of therapeutic efficacy and adverse reactions
Short-term efficacy was evaluated according to the RECIST1.1, with complete response and partial response defined as responders, and stable disease and progressive disease as non-responders. Pathological response was assessed using postoperative specimens based on TRG according to National Comprehensive Cancer Network criteria, with TRG 0-2 classified as responders and TRG 3 as non-responders. All radiological and pathological evaluations were performed by experienced radiologists and pathologists. Treatment efficacy was compared among the three regimens based on RECIST1.1 and TRG assessments. Long-term efficacy was assessed using PFS as the primary endpoint, defined as the time from initiation of chemotherapy to tumor recurrence, development of new GC, or death from any cause, with a follow-up period of 3 years. Overall survival (OS) was defined as the time from initiation of chemotherapy to death from any cause or last follow-up within 3 years. Chemotherapy-related adverse events were evaluated according to the Common Terminology Criteria for Adverse Events, version 5.0.
Statistical analysis
All statistical analyses were performed using R software. Categorical variables were presented as n (%) and compared using the Pearson χ2 test or Fisher’s exact test, as appropriate. A two-sided P value < 0.05 was considered statistically significant. To develop the PFS prediction model, all patients were randomly divided into a training set (70%) and a validation set (30%) with a fixed random seed to ensure reproducibility. In the training set, LASSO regression with 10-fold cross-validation was applied to select candidate variables, and the optimal penalty parameter (λ) was determined based on the λ.1se criterion to reduce overfitting and enhance model stability; a fixed random seed was also applied during this process. Variables with non-zero coefficients were subsequently entered into a multivariable Cox proportional hazards model to estimate hazard ratios (HRs) and 95%CI, thereby identifying independent predictors of PFS. Model performance was evaluated in both the training and validation sets. Discrimination was assessed using receiver operating characteristic (ROC) curves and the area under the curve (AUC), as well as the concordance index (C-index), while calibration was evaluated using calibration plots comparing predicted and observed outcomes.
RESULTS
Basic characteristics of patients
Comparative analysis of baseline characteristics showed no significant differences among the three groups in terms of age, sex, American Society of Anesthesiologists Score classification, tumor differentiation, clinical T stage, lymph node status, Lauren classification, tumor location, RECIST1.1 response, and pre-treatment tumor diameter (P > 0.05), indicating good baseline comparability. Only TRG differed significantly among the groups (P = 0.048; Table 1).
Table 1 Basic patient characteristics and short-term efficacy, n (%).
Short-term efficacy: As shown in Table 1, short-term treatment responses among the three groups were evaluated using RECIST1.1 and TRG. Based on RECIST1.1 criteria, the response rates were 60.7% in the PSOX group, 50.0% in the FOLFOX group, and 67.0% in the SOX + XDL group, with no statistically significant difference among the groups (χ2 = 3.73, P = 0.155), indicating comparable radiological responses. However, TRG-based assessment revealed a significant difference among the three groups (χ2 = 6.08, P = 0.048). The SOX + XDL group demonstrated the highest response rate (84.0%), which was notably higher than that in the PSOX group (72.0%) and the FOLFOX group (69.0%). Overall, although no significant differences were observed in radiological response based on RECIST1.1, the SOX + XDL regimen showed a superior pathological response based on TRG, suggesting a potential advantage in promoting tumor regression.
Survival analysis: Kaplan-Meier analysis was performed to evaluate PFS and OS among the three groups (Figure 1). For PFS, a clear separation of survival curves was observed. The SOX + XDL group consistently demonstrated the highest PFS rate throughout the follow-up period, with its survival curve remaining above those of the PSOX and FOLFOX groups; the PSOX group showed intermediate outcomes, while the FOLFOX group had the lowest PFS. The differences between groups became more pronounced over time, particularly after approximately 18 months. Log-rank testing indicated a statistically significant difference in PFS among the three groups (P = 0.017), suggesting that treatment regimens differed in their ability to delay disease progression. A similar trend was observed for OS. The SOX + XDL group exhibited the most favorable survival outcomes, maintaining consistently higher survival rates than the PSOX and FOLFOX groups throughout follow-up, followed by the PSOX group, with the FOLFOX group showing the poorest survival. The differences further widened over time, especially after 24 months. Log-rank analysis confirmed a significant difference in OS among the three groups (P = 0.002), indicating a substantial impact of treatment strategy on long-term survival. Additionally, risk table and event distribution analyses showed that the SOX + XDL group maintained a higher number of patients at risk and a lower event rate at each time point, further supporting its prognostic advantage. Overall, the SOX + XDL regimen demonstrated superior outcomes in both PFS and OS, followed by PSOX, while FOLFOX was associated with relatively poorer prognosis.
Figure 1 Comparison of progression-free survival and overall survival among the three groups.
A: Progression-free survival; B: Overall survival. FOLFOX: Oxaliplatin plus leucovorin and fluorouracil; OS: Overall survival; PFS: Progression-free survival; PSOX: Plus oxaliplatin and S-1; SOX + XDL: S-1 combined with sintilimab and oxaliplatin.
Adverse reaction
As shown in Table 2, all three treatment regimens were generally well tolerated, and treatment-related adverse events were manageable. There were no statistically significant differences in the incidence of most adverse events among the three groups (P > 0.05). Specifically, the rates of nausea and vomiting, liver toxicity, peripheral neuropathy, neutropenia, fewer white blood cells, fewer platelets, and anemia were comparable across groups, with no significant differences observed (P > 0.05). However, the incidence of alopecia differed significantly among the three groups (χ2 = 17.34, P = 0.002). The PSOX group exhibited a markedly higher rate of alopecia compared with the FOLFOX and SOX + XDL groups, whereas the SOX + XDL group showed a relatively lower incidence. The overall toxicity at levels 3 to 4 was relatively low, mainly involving hematological issues. The most common serious adverse events were anemia (4.70%), neutropenia (3.69%), and leukopenia (3.69%), while grade 3 or higher nausea/vomiting, peripheral sensory neuropathy, and thrombocytopenia were relatively rare. No grade 3 or higher liver toxicity or alopecia was observed, indicating that all three regimens were generally well tolerated. Overall, all three regimens demonstrated acceptable safety profiles, with similar rates of most common adverse events, while the SOX + XDL regimen showed better tolerability in terms of alopecia.
Table 2 Side effects associated with three groups treatment, n (%).
In the training set, LASSO regression was applied to select candidate variables. The λ was determined 10-fold cross-validation, and the final model was selected based on the λ.1se criterion. As λ increased, the regression coefficients of variables gradually shrank toward zero (Figure 2), indicating reduced model complexity. At the optimal λ value, five variables with non-zero coefficients were retained (Figure 3), with detailed coefficients provided in Supplementary Table 1. These variables included poor differentiation, lymph node metastasis, clinical TNM stage IIIC, RECIST1.1, and TRG. Among them, clinical TNM stage IIIC had the largest coefficient (0.477), suggesting the strongest impact on prognosis. Positive lymph node status (0.099) and poor differentiation (0.061) were also positively associated with increased risk of disease progression. In contrast, RECIST1.1 (-0.133) and TRG (-0.101) showed negative coefficients, indicating that better treatment response was associated with a lower risk of progression. Overall, clinical stage and tumor biological characteristics remained key prognostic factors, while treatment response indicators (RECIST1.1 and TRG) also contributed substantially to the predictive model.
Figure 2 Least absolute shrinkage and selection operator regression analysis.
A: Least absolute shrinkage and selection operator regression with 10-fold cross-validation for selection of the optimal penalty parameter; B: Least absolute shrinkage and selection operator coefficient profiles of candidate variables as a function of log(λ).
Figure 3 Least absolute shrinkage and selection operator -selected predictors and their corresponding coefficients.
RECIST: Response Evaluation Criteria in Solid Tumors; TRG: Tumor regression grade.
Cox regression analysis
Based on variables selected by LASSO regression, tumor differentiation, lymph node status, clinical TNM stage, RECIST1.1, and TRG were included in the multivariable Cox proportional hazards model (Table 3). The results showed that poor differentiation was significantly associated with an increased risk of disease progression (HR = 1.86, 95%CI: 1.19-2.91, P = 0.006), whereas no significant difference was observed between moderate and well differentiation (P = 0.340). Patients with positive lymph node metastasis had a higher risk of progression (HR = 1.69, 95%CI: 1.11-2.57, P = 0.013). Clinical stage was also a significant prognostic factor, with stage IIIB (HR = 2.14, 95%CI: 1.39-3.28, P < 0.001) and stage IIIC (HR = 3.94, 95%CI: 2.47-6.28, P < 0.001) identified as adverse predictors of PFS, while no significant difference was observed between stage IIIA and stage IIB (P = 0.199). Regarding treatment response indicators, patients achieving a favorable response based on RECIST1.1 had a significantly lower risk of disease progression (HR = 0.65, 95%CI: 0.46-0.92, P = 0.016), indicating a protective effect. Similarly, lower TRG grades were associated with improved prognosis (HR = 0.56, 95%CI: 0.39-0.82, P = 0.003). Overall, tumor differentiation, lymph node metastasis, and clinical stage were identified as independent adverse prognostic factors, whereas RECIST1.1 response and TRG served as protective factors, jointly influencing PFS in patients with locally advanced GC.
Table 3 Results of multivariate Cox proportional hazards regression analyses.
Based on the independent prognostic factors identified in the multivariable Cox regression analysis (tumor differentiation, lymph node status, clinical TNM stage, RECIST1.1 response, and TRG), a nomogram was developed to predict PFS in patients with locally advanced GC (Figure 4). In this model, each variable was assigned a corresponding score, and the total points were calculated by summing the scores across all variables. The total score was then mapped to an individual linear predictor and the corresponding probabilities of 18-month, 24-month, and 36-month PFS. The nomogram demonstrated that clinical TNM stage contributed the greatest weight to the model, followed by lymph node status and tumor differentiation, highlighting the critical role of tumor stage in prognosis. Meanwhile, treatment response indicators, including RECIST1.1 and TRG, also contributed substantially, indicating their significant predictive value for patient outcomes.
The discriminatory performance of the model was evaluated using time-dependent ROC curve analysis. ROC curves were plotted in both the training and validation sets to assess the predictive accuracy for PFS at different time points (Figure 5). In the training set, the model demonstrated good discrimination across all time points, with an AUC of 0.843 (95%CI: 0.777-0.909) at 18 months, indicating high short-term predictive accuracy; the AUC was 0.810 (95%CI: 0.750-0.869) at 24 months, remaining at a favorable level; and 0.858 (95%CI: 0.754-0.963) at 36 months, suggesting strong long-term predictive performance. In the validation set, the model showed similarly stable performance, with AUCs of 0.837 (95%CI: 0.740-0.933) and 0.835 (95%CI: 0.747-0.922) at 18 months and 24 months, respectively, comparable to those in the training set, indicating good reproducibility and stability. At 36 months, the AUC was 0.748 (95%CI: 0.609-0.887), slightly lower than that in the training set but still within an acceptable range. Overall, the model consistently achieved AUC values above 0.75 in both cohorts, with particularly strong performance at 18 months and 24 months, highlighting its accuracy in short-term to mid-term PFS prediction. The concordance between the training and validation sets further supports its robustness, generalizability, and potential clinical utility.
Figure 5 Progression-free survival receiver operating characteristic curve.
A: 18-month, 24-month, 36-month training set; B: 18-month, 24-month, 36-month validation set. AUC: Area under the curve.
Calibration curves were plotted to assess the agreement between predicted probabilities and observed outcomes for PFS at 18 months, 24 months, and 36 months in both the training and validation sets (Figure 6). In the training set, the model demonstrated good calibration across all time points. At 18 months, the calibration curve closely aligned with the ideal diagonal line, indicating excellent agreement (Brier score: 12.6%). Similarly, good calibration was observed at 24 months (Brier score: 18.0%) and 36 months (Brier score: 14.5%). In the validation set, the model also showed stable calibration performance. The 18-month calibration curve closely approximated the reference line (Brier score: 17.0%), and good agreement was maintained at 24 months (Brier score: 17.7%). Although a slight deviation was observed at 36 months, the overall trend remained consistent with the ideal line (Brier score: 19.3%). Overall, the calibration curves in both the training and validation sets were close to the ideal diagonal, indicating good agreement between predicted and observed outcomes. The relatively low Brier scores across all time points further support the accuracy and stability of the model.
Figure 6 Calibration curve for progression-free survival.
A: 18-month training set; B: 24-month training set; C: 36-month training set; D: 18-month validation set; E: 24-month validation set; F: 36-month validation set.
Decision curve analysis was performed to evaluate the clinical utility of the model at 18 months, 24 months, and 36 months in both the training and validation sets (Figure 7). In the training set, the model demonstrated a consistently higher net benefit across a wide range of threshold probabilities at all time points. At 18 months, the model curve remained above both the “treat-all” and “treat-none” reference strategies, indicating a clear advantage in clinical decision-making. Similar trends were observed at 24 months and 36 months, where the model provided greater net benefit across most threshold probability ranges, suggesting stable clinical applicability over time. In the validation set, the model also showed good clinical usefulness. At 18 months and 24 months, the model curve consistently outperformed the reference strategies across a broad range of thresholds, indicating robust performance in an independent dataset. Although slight fluctuations in net benefit were observed at 36 months, the model still generally outperformed the “treat-all” and “treat-none” approaches.
Figure 7 Decision curve analysis curves of progression-free survival.
A: 18-month training set; B: 24-month training set; C: 36-month training set; D: 18-month validation set; E: 24-month validation set; F: 36-month validation set.
DISCUSSION
This single-center, real-world retrospective cohort study included 298 patients with locally advanced, resectable gastric adenocarcinoma (stage IIB-IIIC) who underwent R0 resection and D2 gastrectomy and compared three treatment strategies: (1) PSOX; (2) FOLFOX; and (3) SOX + XDL. In terms of short-term efficacy, no statistically significant differences were observed among the three groups in terms of RECIST1.1 response rates; however, the response rate was significantly greater in the SOX + XDL group (P = 0.048). With respect to long-term outcomes, Kaplan-Meier analysis revealed that compared with the other groups, the SOX + XDL group achieved superior PFS and OS (log-rank: P = 0.017 for PFS and P = 0.002 for OS). Overall safety was acceptable across all regimens, although the incidence of alopecia differed significantly among the groups (P = 0.002). For prognostic modeling, LASSO regression was applied for variable selection, followed by the construction of a multivariable Cox model incorporating tumor differentiation, lymph node status (clinical N stage), clinical TNM stage, RECIST1.1 response, and TRG. The resulting nomogram demonstrated favorable predictive performance, with relatively high AUC values for 18-month, 24-month, and 36-month PFS in both the training and validation sets; however, the decline in the AUC at 36 months in the validation set suggests some uncertainty in long-term extrapolation. Internationally, triplet perioperative chemotherapy regimens represented by fluorouracil plus leucovorin, oxaliplatin and docetaxel have been established as a standard of care, whereas in East Asia, doublet regimens such as SOX and FOLFOX remain widely used, with evidence supporting their feasibility and efficacy. Notably, some randomized studies have demonstrated the noninferiority and potential interchangeability of SOX compared with FOLFOX, further supporting the clinical relevance of the present findings[12,13]. Because the treatment group was not included in the multivariable Cox model, the independent effect of treatment on prognosis requires further adjusted analysis.
In terms of short-term efficacy, no significant differences in response rates were observed among the three groups based on RECIST1.1, whereas the TRG-defined response rate was significantly greater in the SOX + XDL group. Notably, the clinical TNM stage has the heaviest weight in our model, surpassing RECIST1.1 and TRG. This is clinically reasonable because the cTNM stage reflects the overall baseline tumor burden and disease extent, whereas RECIST1.1 and TRG mainly indicate treatment sensitivity after neoadjuvant therapy. Therefore, treatment response can improve prognosis but cannot completely replace the initial stage. Clinically, these findings suggest that the prognostic assessment of locally advanced GC should integrate baseline staging and post-treatment response to improve risk stratification and guide postoperative monitoring and management. This discrepancy reflects the known difficulty of evaluating neoadjuvant immunotherapy. Immunochemotherapy may induce both tumor cell clearance and inflammatory infiltration, resulting in inconsistency between radiological tumor shrinkage and the actual pathological response[14,15]. In contrast, pathological indicators such as pathological complete response, major pathological response, and TRG more directly reflect tumor eradication and are therefore more sensitive in this setting[16]. In addition, atypical response patterns such as pseudoprogression further limit the value of RECIST alone[17]. In our study, the superior TRG response in the SOX + XDL group is consistent with previous reports showing that SOX plus PD-1 inhibitors can improve the pathological response, particularly in selected subgroups, such as patients with high programmed death-ligand 1 combined positive score, Epstein-Barr virus positivity, or deficient mismatch repair[18]. Similarly, phase II studies of neoadjuvant sintilimab combined with chemotherapy in resectable gastric and gastroesophageal junction cancer have reported encouraging pathological responses with manageable safety, and pathological complete response has been associated with better subsequent survival outcomes[19,20].
In terms of long-term outcomes, this study demonstrated that compared with the PSOX and FOLFOX regimens, the SOX + XDL regimen was associated with superior PFS and OS. These findings are directionally consistent with randomized evidence from advanced or unresectable GC, where XDL combined with chemotherapy has been shown to improve survival. In the ORIENT-16 trial, which was conducted in Chinese patients with unresectable locally advanced, recurrent, or metastatic gastric and gastroesophageal junction adenocarcinoma, XDL plus chemotherapy significantly improved OS compared with placebo plus chemotherapy[18]. These observations align with those of multiple recent large-scale clinical studies indicating that PD-1 inhibitors combined with chemotherapy can significantly improve survival outcomes in patients with locally advanced GC. The underlying mechanism may involve not only the direct cytotoxic effects of chemotherapy but also its ability to induce immunogenic cell death, leading to the release of tumor antigens and the enhancement of antitumor immune responses, thereby producing a synergistic effect with immunotherapy[21,22]. The global phase III CheckMate 649 trial demonstrated that another PD-1 inhibitor, nivolumab, combined with fluoropyrimidine-platinum chemotherapy, significantly improved both OS and PFS compared with chemotherapy alone, further supporting the survival advantage of PD-1-based chemoimmunotherapy in patients with GC[6]. Recent mechanistic evidence has further shown that chemotherapy-induced immunogenic cell death can enhance dendritic cell activation, tumor antigen presentation, and downstream cytotoxic T-cell priming, thereby creating a more permissive immune microenvironment for PD-1 blockade[23]. Although ORIENT-16 was conducted in the first-line treatment setting for advanced disease rather than in the neoadjuvant setting for resectable tumors, it provides important evidence supporting the clinical synergy between XDL and fluoropyrimidine–platinum-based chemotherapy[21]. The separation of survival curves observed in the present study further supports the biological plausibility and clinical relevance of this combination strategy.
In terms of safety, all three regimens demonstrated acceptable tolerability, with most adverse events being manageable grade 1-2 toxicities. With the exception of alopecia, the incidence of common adverse events did not differ significantly among the groups. The overall toxicity profiles – including nausea and vomiting, liver function abnormalities, peripheral sensory neuropathy, myelosuppression, and anemia – were comparable across regimens. Notably, neither the oxaliplatin-fluoropyrimidine-based doublet regimens nor the addition of XDL to the SOX backbone resulted in an unacceptable increase in toxicity. The primary difference among the groups was a greater incidence of alopecia in the PSOX group, which is consistent with the known pharmacological characteristics of nab-paclitaxel. Importantly, compared with conventional chemotherapy, the SOX + XDL regimen did not result in a higher overall incidence of adverse events, which is consistent with the findings of previous reports. In prior studies, the most common hematologic toxicities included leukopenia, neutropenia, and anemia, whereas the most frequent nonhematologic toxicities were elevated transaminases, vomiting, and pneumonia[24-27]. These findings are largely consistent with those of the present study, in which myelosuppression and gastrointestinal reactions were the predominant adverse events and were generally manageable.
XDL is a fully human immunoglobulin G4 monoclonal antibody targeting PD-1 that restores antitumor immunity by blocking the interaction between PD-1 and programmed cell-death protein 1 ligand 1 (PD-L1)/PD-L2[28,29]. PD-1 is an important inhibitory receptor on T cells, and its activation suppresses T-cell function by attenuating costimulatory signaling[30,31]. Experimental studies have shown that this inhibitory effect is closely related to the disruption of the CD28 pathway and downstream signaling molecules such as SHP2, which together contribute to immune suppression and T-cell exhaustion[32,33]. These mechanisms provide a biological basis for the favorable efficacy of PD-1 blockade, including XDL, in advanced GC.
Several limitations of this study should be acknowledged. First, as a single-center retrospective analysis, it is inherently subject to potential selection and information biases. Although baseline characteristics were generally comparable among the three groups, residual confounding cannot be fully excluded. Second, the sample sizes across treatment groups were imbalanced, particularly with a relatively small number of patients in the FOLFOX group, which may have affected the stability of the statistical analyses and the power of the intergroup comparisons. Third, key immunological biomarkers – such as PD-L1 expression, microsatellite instability status, tumor mutational burden, and tumor immune microenvironment-related indicators – were not included. These factors are known to be closely associated with the efficacy of immunotherapy, and their absence may limit the mechanistic interpretation of the observed differences in treatment outcomes as well as the predictive accuracy of the model in immunotherapy-treated populations. In addition, the prediction model was primarily based on clinicopathological variables and treatment response indicators and did not incorporate molecular subtyping or multiomics data, which may have limited its predictive performance and external applicability. Although internal validation was performed using training and validation sets, external validation in an independent cohort is still lacking, and the generalizability of the model requires further evaluation. Additionally, because the cTNM stage already includes nodal status, including both the cN stage and the cTNM stage in a multivariable model may introduce redundancy and potential multicollinearity. Therefore, the estimated effects of these two variables should be interpreted cautiously. Finally, the follow-up duration was relatively limited, and some patients had not yet reached endpoint events, which may have affected the assessment of long-term survival outcomes. Future studies with multicenter, large-scale, and prospective designs that incorporate immunological biomarkers and molecular characteristics are warranted to further validate and optimize both therapeutic strategies and predictive models.
CONCLUSION
In summary, the results of the present study suggest that, compared with PSOX and FOLFOX, the SOX + XDL regimen may be associated with greater pathological response rates in the neoadjuvant treatment of locally advanced GC and may be associated with superior PFS and OS without a significant increase in severe adverse events, indicating a favorable and manageable safety profile. In addition, the predictive model developed on the basis of LASSO regression and multivariable Cox analysis demonstrated good discrimination and calibration, suggesting that integrating clinicopathological features with treatment response indicators can improve the accuracy of prognostic assessment. These findings suggest that the SOX + XDL regimen may represent a promising perioperative treatment option with a balanced efficacy and safety profile and provide supportive evidence for post-treatment prognostic assessment and risk stratification in patients with locally advanced GC.
Yu J, Gao Y, Chen L, Wu D, Shen Q, Zhao Z, Liu W, Yang H, Zhang Q, Wang X, Hu P, Zheng Z, Wang X, Liu H, Xu Z, Yan Z, Wu Y, Jin M, Zhang Q, Liu X, Zhu K, Shou C. Effect of S-1 Plus Oxaliplatin Compared With Fluorouracil, Leucovorin Plus Oxaliplatin as Perioperative Chemotherapy for Locally Advanced, Resectable Gastric Cancer: A Randomized Clinical Trial.JAMA Netw Open. 2022;5:e220426.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 26][Cited by in RCA: 24][Article Influence: 6.0][Reference Citation Analysis (0)]
Janjigian YY, Shitara K, Moehler M, Garrido M, Salman P, Shen L, Wyrwicz L, Yamaguchi K, Skoczylas T, Campos Bragagnoli A, Liu T, Schenker M, Yanez P, Tehfe M, Kowalyszyn R, Karamouzis MV, Bruges R, Zander T, Pazo-Cid R, Hitre E, Feeney K, Cleary JM, Poulart V, Cullen D, Lei M, Xiao H, Kondo K, Li M, Ajani JA. First-line nivolumab plus chemotherapy versus chemotherapy alone for advanced gastric, gastro-oesophageal junction, and oesophageal adenocarcinoma (CheckMate 649): a randomised, open-label, phase 3 trial.Lancet. 2021;398:27-40.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 2501][Cited by in RCA: 2434][Article Influence: 486.8][Reference Citation Analysis (9)]
Sato S, Kunisaki C, Tanaka Y, Sato K, Miyamoto H, Yukawa N, Fujii Y, Kimura J, Takagawa R, Takahashi M, Kosaka T, Akiyama H, Saigusa Y, Taguri M, Yamanaka T, Endo I. A Phase II Study of Tri-weekly Low-dose Nab-paclitaxel Chemotherapy for Patients with Advanced Gastric Cancer.Anticancer Res. 2018;38:6911-6917.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 11][Cited by in RCA: 14][Article Influence: 1.8][Reference Citation Analysis (0)]
Al-Batran SE, Homann N, Pauligk C, Goetze TO, Meiler J, Kasper S, Kopp HG, Mayer F, Haag GM, Luley K, Lindig U, Schmiegel W, Pohl M, Stoehlmacher J, Folprecht G, Probst S, Prasnikar N, Fischbach W, Mahlberg R, Trojan J, Koenigsmann M, Martens UM, Thuss-Patience P, Egger M, Block A, Heinemann V, Illerhaus G, Moehler M, Schenk M, Kullmann F, Behringer DM, Heike M, Pink D, Teschendorf C, Löhr C, Bernhard H, Schuch G, Rethwisch V, von Weikersthal LF, Hartmann JT, Kneba M, Daum S, Schulmann K, Weniger J, Belle S, Gaiser T, Oduncu FS, Güntner M, Hozaeel W, Reichart A, Jäger E, Kraus T, Mönig S, Bechstein WO, Schuler M, Schmalenberg H, Hofheinz RD; FLOT4-AIO Investigators. Perioperative chemotherapy with fluorouracil plus leucovorin, oxaliplatin, and docetaxel versus fluorouracil or capecitabine plus cisplatin and epirubicin for locally advanced, resectable gastric or gastro-oesophageal junction adenocarcinoma (FLOT4): a randomised, phase 2/3 trial.Lancet. 2019;393:1948-1957.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 2158][Cited by in RCA: 1994][Article Influence: 284.9][Reference Citation Analysis (11)]
Zhang X, Liang H, Li Z, Xue Y, Wang Y, Zhou Z, Yu J, Bu Z, Chen L, Du Y, Wang X, Wu A, Li G, Su X, Xiao G, Cui M, Wu D, Chen L, Wu X, Zhou Y, Zhang L, Dang C, He Y, Zhang Z, Sun Y, Li Y, Chen H, Bai Y, Qi C, Yu P, Zhu G, Suo J, Jia B, Li L, Huang C, Li F, Ye Y, Xu H, Wang X, Yuan Y, E JY, Ying X, Yao C, Shen L, Ji J; RESOLVE study group. Perioperative or postoperative adjuvant oxaliplatin with S-1 versus adjuvant oxaliplatin with capecitabine in patients with locally advanced gastric or gastro-oesophageal junction adenocarcinoma undergoing D2 gastrectomy (RESOLVE): an open-label, superiority and non-inferiority, phase 3 randomised controlled trial.Lancet Oncol. 2021;22:1081-1092.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 375][Cited by in RCA: 342][Article Influence: 68.4][Reference Citation Analysis (4)]
Smyth EC, Fassan M, Cunningham D, Allum WH, Okines AF, Lampis A, Hahne JC, Rugge M, Peckitt C, Nankivell M, Langley R, Ghidini M, Braconi C, Wotherspoon A, Grabsch HI, Valeri N. Effect of Pathologic Tumor Response and Nodal Status on Survival in the Medical Research Council Adjuvant Gastric Infusional Chemotherapy Trial.J Clin Oncol. 2016;34:2721-2727.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 242][Cited by in RCA: 247][Article Influence: 24.7][Reference Citation Analysis (3)]
Seymour L, Bogaerts J, Perrone A, Ford R, Schwartz LH, Mandrekar S, Lin NU, Litière S, Dancey J, Chen A, Hodi FS, Therasse P, Hoekstra OS, Shankar LK, Wolchok JD, Ballinger M, Caramella C, de Vries EGE; RECIST working group. iRECIST: guidelines for response criteria for use in trials testing immunotherapeutics.Lancet Oncol. 2017;18:e143-e152.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 2034][Cited by in RCA: 1959][Article Influence: 217.7][Reference Citation Analysis (6)]
Jiang H, Yu X, Li N, Kong M, Ma Z, Zhou D, Wang W, Wang H, Wang H, He K, Li Z, Lu Y, Zhang J, Zhao K, Zhang Y, Xu N, Li Z, Liu Y, Wang Y, Wang Y, Teng L. Efficacy and safety of neoadjuvant sintilimab, oxaliplatin and capecitabine in patients with locally advanced, resectable gastric or gastroesophageal junction adenocarcinoma: early results of a phase 2 study.J Immunother Cancer. 2022;10:e003635.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 94][Cited by in RCA: 95][Article Influence: 23.8][Reference Citation Analysis (0)]
Beshr MS, Beshr IA, Al Hayek M, Alfaqaih SM, Abuajamieh M, Basheer E, Wali AK, Ekreer M, Chenfouh I, Khashan A, Hassan ET, Elnaami SM, Elhadi M. PD-1/PD-L1 Inhibitors in Combination With Chemo or as Monotherapy vs. Chemotherapy Alone in Advanced, Unresectable HER2-Negative Gastric, Gastroesophageal Junction, and Esophageal Adenocarcinoma: A Meta-Analysis.Clin Oncol (R Coll Radiol). 2024;36:797-808.
[RCA] [PubMed] [DOI] [Full Text][Cited by in RCA: 12][Reference Citation Analysis (0)]
Li N, Li Z, Fu Q, Zhang B, Zhang J, Wan XB, Lu CM, Wang JB, Deng WY, Ma YJ, Bie LY, Wang MY, Li J, Xia QX, Wei C, Luo SX. Efficacy and safety of neoadjuvant sintilimab in combination with FLOT chemotherapy in patients with HER2-negative locally advanced gastric or gastroesophageal junction adenocarcinoma: an investigator-initiated, single-arm, open-label, phase II study.Int J Surg. 2024;110:2071-2084.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 28][Cited by in RCA: 31][Article Influence: 15.5][Reference Citation Analysis (1)]
Shitara K, Takashima A, Fujitani K, Koeda K, Hara H, Nakayama N, Hironaka S, Nishikawa K, Makari Y, Amagai K, Ueda S, Yoshida K, Shimodaira H, Nishina T, Tsuda M, Kurokawa Y, Tamura T, Sasaki Y, Morita S, Koizumi W. Nab-paclitaxel versus solvent-based paclitaxel in patients with previously treated advanced gastric cancer (ABSOLUTE): an open-label, randomised, non-inferiority, phase 3 trial.Lancet Gastroenterol Hepatol. 2017;2:277-287.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 173][Cited by in RCA: 159][Article Influence: 17.7][Reference Citation Analysis (3)]
