Zhou DX, Jiang SW, Zhang CG, Cai BJ, Lv HD. Perioperative SOX plus sintilimab vs P-SOX and CAPOX in advanced gastric cancer: A real-world comparison. World J Gastroenterol 2026; 32(29): 119809 [DOI: 10.3748/wjg.119809]
Corresponding Author of This Article
Hai-Dong Lv, Professor, Department of Gastrointestinal Cancer Surgery, Qinghai Provincial People’s Hospital, No. 2 Gonghe Road, Chengdong District, Xining 810000, Qinghai Province, China. lvhaid@126.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
Zhou DX, Jiang SW, Zhang CG, Cai BJ, Lv HD. Perioperative SOX plus sintilimab vs P-SOX and CAPOX in advanced gastric cancer: A real-world comparison. World J Gastroenterol 2026; 32(29): 119809 [DOI: 10.3748/wjg.119809]
Author contributions: Zhou DX and Jiang SW were responsible for study conception and design, data analysis, and manuscript drafting, contributed equally as co-first authors; Zhang CG and Cai BJ contributed to data collection and interpretation; Lv HD supervised the study and revised the manuscript critically for important intellectual content. All authors approved the final version of the manuscript.
AI contribution statement: DeepSeek(V3) was used during manuscript preparation. DeepSeek was used only for language polishing and translation assistance. It was not used for data analysis, statistical analysis, or scientific writing assistance.
Institutional review board statement: This study was approved by the Ethics Committee of the Qinghai University Affiliated Hospital, No. SL-2022-035.
Informed consent statement: Written informed consent was obtained from all participants.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
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: Hai-Dong Lv, Professor, Department of Gastrointestinal Cancer Surgery, Qinghai Provincial People’s Hospital, No. 2 Gonghe Road, Chengdong District, Xining 810000, Qinghai Province, China. lvhaid@126.com
Received: February 6, 2026 Revised: February 25, 2026 Accepted: March 27, 2026 Published online: August 7, 2026 Processing time: 161 Days and 14.7 Hours
Abstract
BACKGROUND
Advanced gastric cancer (GC) remains a major cause of cancer-related mortality worldwide. Perioperative chemotherapy has been established as a standard treatment strategy to improve resectability and survival outcomes. Recently, the incorporation of immune checkpoint inhibitors into chemotherapy regimens has demonstrated promising survival benefits in advanced and metastatic settings. However, real-world evidence comparing immunochemotherapy with intensified chemotherapy regimens in the perioperative context remains limited, and the prognostic factors influencing progression-free survival (PFS) in this population are not fully understood.
AIM
To compare the efficacy and safety of oxaliplatin plus S-1 combined with sintilimab (SOX + XDL), nab-paclitaxel plus oxaliplatin and S-1 (P-SOX), and capecitabine plus oxaliplatin (CAPOX) as perioperative treatment regimens for patients with advanced GC.
METHODS
A total of 323 patients with advanced GC who received neoadjuvant chemotherapy followed by standard D2 radical gastrectomy were retrospectively analyzed. Among them, 101 patients were treated with SOX + XDL, 148 with P-SOX, and 74 with CAPOX. By using the sample function in R software, all patients were randomly assigned to a training set and a validation set, which were used for model development and validation, respectively. Short-term efficacy, long-term outcomes, and treatment-related adverse events were compared among the three groups. In addition, clinical factors associated with PFS were further investigated.
RESULTS
Based on the tumor regression grade, compared with the P-SOX and CAPOX groups, the SOX + XDL group achieved a significantly greater pathological response (85.15% vs 72.30% vs 59.46%, P < 0.001). According to the Response Evaluation Criteria in Solid Tumors version 1.1 criteria, the objective response rate was numerically higher in the SOX + XDL group (68.32%) than in the P-SOX (60.14%) and CAPOX (55.41%) groups, but the difference was not statistically significant (P = 0.196). Kaplan-Meier analysis revealed significant differences in overall survival (OS) and PFS among the three regimens. Patients treated with SOX + XDL had better OS (log-rank P = 0.004) and longer PFS (log-rank P = 0.026) than did those treated with P-SOX or CAPOX. All regimens were generally well tolerated, with predominantly mild to moderate adverse events. No significant differences in hematologic toxicity or liver function abnormalities were detected among the groups (all P > 0.05). However, alopecia occurred more frequently in the P-SOX and CAPOX groups than in the SOX + XDL group (P = 0.002).
CONCLUSION
Compared with the other regimens, the SOX + XDL regimen showed superior short-term efficacy and improved survival outcomes, with better OS and PFS. Favorable treatment response, smaller tumor size (≤ 2 cm), earlier clinical tumor-node-metastasis stage, and well-differentiated tumors were associated with better prognosis in patients with advanced GC.
Core Tip: This real-world study systematically compared perioperative oxaliplatin plus S-1 combined with sintilimab, nab-paclitaxel plus oxaliplatin and S-1, and capecitabine plus oxaliplatin regimens in patients with advanced gastric cancer, focusing on treatment efficacy, survival outcomes, and safety. Beyond conventional survival analyses, a prognostic model was developed and validated to stratify progression-free survival risk using readily available clinicopathological and treatment response variables. The findings demonstrate that oxaliplatin plus S-1 combined with sintilimab is associated with favorable survival outcomes with manageable toxicity, and the proposed nomogram provides a practical tool for individualized risk assessment and treatment decision-making in perioperative advanced gastric cancer.
Citation: Zhou DX, Jiang SW, Zhang CG, Cai BJ, Lv HD. Perioperative SOX plus sintilimab vs P-SOX and CAPOX in advanced gastric cancer: A real-world comparison. World J Gastroenterol 2026; 32(29): 119809
Gastric cancer (GC) remains one of the leading causes of cancer-related mortality worldwide, and a considerable proportion of patients are diagnosed at an advanced stage[1,2]. For patients with locally advanced GC, perioperative chemotherapy has become the standard treatment strategy, aiming to improve tumor resectability and long-term survival outcomes. Platinum-based chemotherapy regimens, owing to their proven efficacy and manageable toxicity profiles, have been widely adopted in East Asian populations[3-5]. However, despite continuous optimization of perioperative treatment strategies, the overall prognosis of patients with advanced GC remains unsatisfactory[6], underscoring an urgent need to explore more effective therapeutic approaches. With the rapid advances in immunotherapy and molecular profiling, systemic treatment strategies for GC have undergone substantial evolution in recent years. The 2025 National Comprehensive Cancer Network (NCCN) guidelines have incorporated immune checkpoint inhibitors combined with platinum-based chemotherapy as an important first-line treatment option for advanced or metastatic GC, particularly in biomarker-selected populations. These guidelines emphasize stratified decision-making based on biomarkers such as human epidermal growth factor receptor 2 (HER2), programmed cell death ligand 1 (PD-L1), microsatellite instability/deficient mismatch repair, and claudin 18.2 (CLDN18.2) to optimize patient selection and maximize therapeutic benefit[7]. Similarly, the European Society for Medical Oncology living guidelines recommend programmed cell death protein 1 (PD-1) inhibitors in combination with chemotherapy as first-line therapy for HER2-negative patients with PD-L1 combined positive scores (CPS) above predefined thresholds, and continuously update treatment pathways according to emerging biomarker evidence. In addition to immunotherapy, treatments targeting novel molecular markers have also been incorporated into contemporary guidelines. Monoclonal antibodies targeting CLDN18.2, such as zolbetuximab, have demonstrated significant survival benefits when combined with platinum-based chemotherapy in patients with CLDN18.2-positive and HER2-negative GC. Accordingly, European Society for Medical Oncology has issued specific first-line treatment recommendations for this biomarker-defined subgroup, positioning “targeted therapy plus chemotherapy” as an important strategy alongside immunotherapy-based combinations[8,9]. Within the Chinese guideline framework, the accumulation of evidence supporting domestically developed immunotherapeutic agents has further expanded treatment options. The 2025 Chinese Society of Clinical Oncology guidelines have incorporated cadonilimab, a PD-1/cytotoxic T lymphocyte-associated antigen-4 bispecific antibody, in combination with chemotherapy into the recommended treatment framework, indicating that the selection of immunochemotherapy regimens in perioperative and real-world clinical settings is becoming increasingly diverse and refined[10,11]. Against this background, a systematic evaluation of the efficacy of different chemotherapy regimens combined with immunotherapy in the perioperative setting, together with validation using pathological response and survival outcomes, is of considerable clinical importance for guiding individualized treatment strategies in advanced GC[12,13]. However, direct comparative evidence regarding the efficacy of immunochemotherapy regimens vs intensified chemotherapy strategies remains limited. Therefore, the present study was conducted to compare the efficacy and safety of oxaliplatin plus S-1 combined with sintilimab (SOX + XDL) with nab-paclitaxel plus oxaliplatin and S-1 (P-SOX) and capecitabine plus oxaliplatin (CAPOX) alone as perioperative treatment strategies for patients with advanced GC, with the aim of providing evidence to optimize clinical decision-making.
MATERIALS AND METHODS
Patient information
Patients with advanced GC treated at the Department of Gastrointestinal Oncology, Qinghai University Affiliated Hospital between January 2017 and April 2022 were retrospectively collected. 23 were excluded due to incomplete baseline clinicopathological data. A total of 323 eligible patients were included in this study, of whom 101 were treated with SOX + XDL, 148 with P-SOX, and 74 with CAPOX. Using the sample function in R software, all patients were randomly assigned in a 7:3 ratio to a training set and a validation set, which were used for model development and validation, respectively. All patients received perioperative chemotherapy consisting of 2-4 cycles before surgery and 2-4 cycles after surgery, for a total of 6-8 cycles, followed by standard D2 radical gastrectomy. Postoperatively, patients were followed up regularly for up to 3 years or until the occurrence of study endpoints. The specific treatment plan can be found in Supplementary material.
Inclusion criteria and exclusion criteria
Patients were eligible for inclusion if they met all of the following 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 tumor-node-metastasis (TNM) staging system of the International Union Against Cancer, with no evidence of distant organ metastasis, no concomitant malignancies, and achievement of R0 resection (no macroscopic or microscopic residual tumor); (2) Receipt of perioperative chemotherapy, including 2-4 cycles of neoadjuvant chemotherapy and 2-4 cycles of adjuvant chemotherapy (total of 6-8 cycles), with all surgical procedures performed at our institution in accordance with NCCN and Chinese Society of Clinical Oncology guidelines; (3) Primary tumor size measurable by computed tomography or magnetic resonance imaging, with postoperative pathological examination available; and (4) Eastern Cooperative Oncology Group performance status ≤ 1, adequate tolerance to chemotherapy, and acceptable hepatic, renal, hematologic, and cardiopulmonary function. Patients were excluded if they met any of the following conditions: (1) Known hypersensitivity to chemotherapy agents or contraindications to chemotherapy, or the presence of severe comorbid conditions such as active infection, gastrointestinal bleeding, pyloric obstruction, or gastrointestinal perforation; (2) Previous history of other malignancies treated with radiotherapy, chemotherapy, biological therapy, or surgery; or (3) Incomplete clinical data or unavailable imaging data precluding accurate measurement of tumor size.
Efficacy and adverse reactions
(1) Short-term efficacy assessment: Clinical response to chemotherapy was evaluated according to the Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST 1.1). Complete response and partial response were defined as an effective response, whereas stable disease and progressive disease were classified as an ineffective response. Pathological response was assessed based on postoperative pathological examination using the tumor regression grade (TRG) system according to the NCCN criteria, in which TRG grades 0, 1, and 2 were considered effective responses, and TRG grade 3 was defined as an ineffective response. All assessments were performed with the assistance of experienced pathologists and radiologists. Chemotherapeutic efficacy was subsequently compared among the three treatment regimens based on RECIST 1.1 and TRG evaluations; (2) Long-term efficacy assessment: The secondary endpoint was overall survival (OS), defined as the time from initiation of chemotherapy to death from any cause or the date of last follow-up, with a maximum follow-up duration of 3 years. The primary endpoint was progression-free survival (PFS), defined as the time from the start of chemotherapy to primary GC recurrence, development of new GC, or death from any cause, whichever occurred first, with a follow-up period of up to 3 years; and (3) Assessment of chemotherapy-related adverse events: Chemotherapy-related adverse events were evaluated and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 5.0.
Statistical analysis
All statistical analyses were performed using R software (version 4.4.1). Continuous variables were first assessed for normality. Variables with a normal distribution were expressed as mean ± SD and compared using the independent-samples t test for two-group comparisons or one-way analysis of variance for multiple-group comparisons. Variables with a non-normal distribution were presented as median (interquartile range), and group differences were evaluated using nonparametric rank-sum tests. Categorical variables were summarized as counts and percentages and compared using the χ2 test or Fisher’s exact test when the expected cell counts were insufficient. Patients were randomly assigned to a training set (70%) and a validation set (30%) using the sample function in R. The training set was used for model development, whereas the validation set was used for model performance evaluation. In the training set, PFS was defined as the outcome variable. Univariate Cox proportional hazards regression analyses were initially conducted to identify potential predictors (P < 0.10), and variables meeting this criterion were subsequently entered into a multivariate Cox regression model to determine independent prognostic factors (P < 0.05). A nomogram prediction model was constructed based on the final set of independent predictors and internally validated using the bootstrap resampling method. Model calibration was assessed using calibration curves, while discriminative performance was evaluated using time-dependent receiver operating characteristic (ROC) curves to calculate the concordance index (C-index) and area under the curve (AUC). Decision curve analysis (DCA) was performed to assess the clinical net benefit of the model across a range of threshold probabilities, thereby evaluating its clinical utility. The established nomogram was further validated in the validation set for external performance assessment. In this study, nomogram construction and calibration analyses were performed using the rms package, with bootstrap validation implemented within the same framework. Time-dependent ROC analyses were conducted using the riskRegression, ggplot2, and ggprism packages. DCA was performed using the ggDCA package, and nomogram score calculation was carried out using the nomogramFormula package. Survival analyses and determination of optimal cutoff values were conducted using the survminer package. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant.
RESULTS
Basic characteristics of patients
A total of 323 patients with advanced GC were included in this study, including 148 patients in the P-SOX group, 101 patients in the SOX + XDL group, and 74 patients in the CAPOX group. As shown in (Table 1), the baseline clinical characteristics were well balanced among the three groups. No statistically significant differences were observed with respect to age, sex, American Society of Anesthesiologists (ASA) classification, tumor differentiation, tumor location, Lauren classification, preoperative T stage, preoperative N stage, or clinical TNM stage (all P > 0.05).
Table 1 Basic patient characteristics and short-term efficacy, n (%).
Based on TRG assessment, 107 patients (72.30%) in the P-SOX group were classified as responders, while 41 patients (27.70%) were classified as non-responders. In the SOX + XDL group, 86 patients (85.15%) achieved an effective response, whereas 15 patients (14.85%) were deemed ineffective. In the CAPOX group, 44 patients (59.46%) met the criteria for effectiveness, while 30 patients (40.54%) were assessed as ineffective. The distribution of TRG grades differed significantly among the three groups (P < 0.001). According to RECIST version 1.1 criteria, 89 patients (60.14%) in the P-SOX group achieved an objective response (complete response and partial response), while 59 patients (39.86%) did not. The objective response rate was 68.32% in the SOX + XDL group and 55.41% in the CAPOX group. No statistically significant difference was observed among the three groups in terms of RECIST 1.1-based responses (P = 0.196). Overall, the SOX + XDL group regimen demonstrated a higher proportion of objective responses under both TRG and RECIST 1.1 assessments, with a particularly pronounced advantage in pathological response, whereas the imaging-based response rate showed only a favorable trend without reaching statistical significance (Table 1).
Kaplan-Meier survival analysis revealed a significant difference in PFS among the treatment groups (log-rank P = 0.026). During early follow-up, the PFS curves were largely comparable; however, as follow-up progressed, the curves gradually diverged. The SOX + XDL group consistently exhibited a higher probability of PFS throughout the follow-up period, with a more pronounced survival advantage emerging in the mid-to-late follow-up phase (approximately after 18 months). In contrast, the PFS curves of the P-SOX and CAPOX groups declined more steeply, suggesting relatively limited long-term disease control. OS analysis also demonstrated a statistically significant difference among the three treatment strategies (log-rank P = 0.004). With extended follow-up, the OS curves progressively separated, with the SOX + XDL group maintaining a consistently higher OS probability. Notably, during long-term follow-up (approximately 30-36 months), patients in the SOX + XDL group continued to exhibit relatively stable survival outcomes, whereas survival probabilities in the P-SOX and CAPOX groups declined steadily. Collectively, these findings suggest that the SOX + XDL group regimen confers a potential advantage in terms of long-term survival benefit (Figure 1).
Figure 1 Comparison of progression-free survival and overall survival among the three groups.
A: Progression-free survival; B: Overall survival. PSOX: Nab-paclitaxel plus oxaliplatin and S-1; SOX + XDL: Oxaliplatin plus S-1 combined with sintilimab; CAPOX: Capecitabine plus oxaliplatin.
Adverse reaction
As shown in (Table 2), the overall safety profiles of the three treatment regimens were acceptable and manageable, with adverse events predominantly mild to moderate in severity. The most common adverse events included nausea and vomiting, alopecia, peripheral sensory neuropathy, and hematologic toxicities. Regarding gastrointestinal toxicity, nausea and vomiting were frequently observed across all three groups, but most events were graded as 1-2, with a low incidence of grade ≥ 3 reactions. No statistically significant differences were observed among the treatment groups (P = 0.569). In terms of hepatic toxicity, the majority of patients did not experience liver function abnormalities; a small proportion developed grade 1-2 elevations in transaminases, and no severe hepatic impairment was observed. The incidence of liver toxicity did not differ significantly among the three groups (P = 0.966). Peripheral sensory neuropathy occurred in all treatment groups, predominantly at mild to moderate grades, while grade 3 neuropathy was rare. No significant differences were detected among the groups (P = 0.944). Hematologic toxicities, including neutropenia, leukopenia, thrombocytopenia, and anemia, were relatively common but were mostly grade 1-2 in severity, with a low incidence of grade ≥ 3 events. The rates of these hematologic adverse events were comparable among the three regimens (neutropenia P = 0.362; leukopenia P = 0.455; thrombocytopenia P = 0.319; anemia P = 0.376). In contrast, the incidence of alopecia differed significantly among the treatment groups (χ² = 17.08, P = 0.002), with higher rates observed in the P-SOX and CAPOX groups, whereas the SOX + XDL group exhibited a relatively lower incidence. Overall, the three treatment regimens demonstrated similar safety profiles, characterized mainly by mild to moderate toxicities, all of which were within acceptable and clinically manageable ranges.
Table 2 Side effects associated with three groups treatment, n (%).
Single and multiple factor Cox regression analysis
Univariate Cox regression analysis of baseline characteristics and short-term treatment response indicators (RECIST 1.1 and TRG) in the training set patients demonstrated that tumor differentiation, tumor stage, treatment response parameters, and tumor diameter were significantly associated with patient prognosis. Specifically, patients with poorly differentiated tumors, cT4 stage, cN1 stage, and higher clinical TNM stage exhibited a significantly increased risk of disease progression, whereas patients achieving an effective response according to RECIST 1.1 and TRG criteria showed a markedly reduced risk. Variables with potential clinical relevance and those with a P value < 0.10 in the univariate analysis were subsequently entered into the multivariate Cox regression model. The multivariate analysis revealed that poor tumor differentiation, advanced clinical TNM stage, and larger tumor diameter remained independent risk factors for adverse outcomes. In contrast, effective responses as assessed by RECIST 1.1 and TRG continued to serve as independent protective factors after adjustment for potential confounders. Age, ASA classification, sex, cT stage, cN stage, Lauren classification, and tumor location were not significantly associated with outcome events in the multivariate analysis (Table 3).
Table 3 Results of univariate and multivariate Cox proportional hazards regression analyses.
Based on the results of univariate and multivariate Cox proportional hazards regression analyses, five independent risk factors were ultimately identified and incorporated into the construction of a nomogram. In the nomogram, the top scale (“points”) represents the score assigned to each variable. For each risk factor, a vertical line is drawn upward to determine the corresponding point value. The sum of the points for all variables yields the total score, which can be located on the “total points” axis to estimate the probabilities of PFS at 18 months, 2 years, and 3 years. In the corresponding risk stratification, lower total scores indicate a greater potential benefit from P-SOX chemotherapy. Notably, patients with advanced GC who achieved an effective response to perioperative chemotherapy (RECIST 1.1 response and favorable TRG), had a tumor diameter ≤ 2 cm, well-differentiated histology, and cTNM stage IIB demonstrated the highest PFS and derived the greatest therapeutic benefit (Figure 2).
The discriminatory performance of the predictive model was evaluated using time-dependent ROC curves, with the AUC and C-index calculated accordingly. As shown in Figure 3, the model’s discriminative ability at different time points (18 months, 24 months, and 36 months) and its temporal stability were assessed using time-dependent ROC curves and a dynamic C-index, respectively. At the 18-month time point, the model demonstrated good discriminative performance in both the training and validation sets. The AUC was 0.877 [95% confidence interval (CI): 0.819-0.935] in the training set and 0.799 (95%CI: 0.691-0.907) in the validation set. The ROC curves were clearly above the diagonal line, indicating strong discriminative ability for short-term outcome prediction. At 24 months, the predictive performance showed a slight decline compared with that at 18 months but remained acceptable. The AUC was 0.839 (95%CI: 0.784-0.893) in the training set and 0.766 (95%CI: 0.670-0.861) in the validation set. The ROC curves exhibited similar trends between the two sets, suggesting reasonable stability of the model during mid-term follow-up. At 36 months, the model continued to demonstrate predictive value. The AUC was 0.819 (95%CI: 0.763-0.875) in the training set and 0.762 (95%CI: 0.661-0.863) in the validation set. Although a gradual decline in AUC was observed with longer follow-up, the overall discriminative ability remained superior to random prediction. The dynamic C-index further evaluated changes in model performance over time. The C-index remained consistently above 0.70 in both datasets, with a value of 0.777 (95%CI: 0.738-0.816) in the training set and 0.718 (95%CI: 0.648-0.788) in the validation set. Overall, the model maintained relatively stable predictive performance throughout the follow-up period, without evidence of substantial performance deterioration.
Figure 3 Progression-free survival receiver operating characteristic curve.
A: 18-month training and validation set; B: 24-month training and validation set; C: 36-month training and validation set; D: Concordance index over time.
Figure 4 presents the combined calibration curves of the nomogram model in the training and validation sets, evaluated using bootstrap resampling. At the 18-month time point, the calibration curves in the training set (Figure 4A) and validation set (Figure 4D) demonstrated close agreement between the predicted and observed survival probabilities, particularly when the predicted probabilities ranged from 0.60 to 0.90, with minimal deviation from the ideal 45° reference line. Although wider confidence intervals were observed in the lower predicted probability range (approximately 0.20-0.40), the point estimates remained close to the ideal line, indicating good calibration performance for short-term survival prediction. At the 24-month time point (Figure 4B and E), the model maintained satisfactory calibration consistency in both the training and validation sets. When predicted probabilities were within the range of 0.40-0.80, the observed survival probabilities closely matched the predicted values, with only minor deviations in the intermediate probability range. Overall, the calibration curves remained well aligned with the ideal reference line. At the 36-month time point (Figure 4C and F), as the follow-up period extended, the confidence intervals in the low to intermediate predicted probability range (approximately 0.10-0.40) became wider, suggesting increased uncertainty in long-term prediction. Nevertheless, in the moderate to high predicted probability range (approximately 0.50-0.80), the predicted survival probabilities continued to show good agreement with the observed outcomes. These findings indicate that the nomogram model retains acceptable calibration performance for long-term outcome prediction despite increased uncertainty at lower probability levels.
Figure 4 Calibration curve for progression-free survival.
A-C: Training set (18 months, 24 months, 36 months); D-F: Validation set (18 months, 24 months, 36 months).
Furthermore, the clinical decision-making value of the prediction model was evaluated using DCA curves in Figure 5, which compared the net clinical benefit of the model with the “treat-all” and “treat-none” strategies at different time points (18 months, 24 months, and 36 months) in both the training and validation sets. At the 18-month time point, in both the training set (Figure 5A) and validation set (Figure 5D), the model consistently yielded higher net benefit than the “treat-all” and “treat-none” strategies across a threshold probability range of approximately 0.05-0.60, indicating superior clinical utility within this range. When the threshold probability exceeded 0.60, the net benefit of the model gradually declined and approached zero, suggesting diminished clinical usefulness at higher risk thresholds. At the 24-month time point, the model curves in the training set (Figure 5B) and validation set (Figure 5E) remained above both reference strategies across a threshold probability range of approximately 0.10-0.70. Notably, the model demonstrated relatively stable and superior net benefit within the moderate threshold range (approximately 0.20-0.50). As the threshold probability increased beyond 0.70, the net benefit progressively decreased and approached zero. At the 36-month time point, the model in both the training set (Figure 5C) and validation set (Figure 5F) continued to show a clear net benefit advantage over a broad threshold probability range (approximately 0.10-0.65), indicating sustained clinical usefulness for long-term outcome prediction. However, at higher threshold probabilities (> 0.65-0.70), the net benefit declined markedly, suggesting limited clinical benefit when more stringent risk thresholds are applied. Overall, the model consistently provided greater net clinical benefit than the “treat-all” and “treat-none” strategies across reasonable threshold probability ranges at 18 months, 24 months, and 36 months in both the training and validation sets. These findings support the potential clinical applicability of the nomogram model across different follow-up periods.
Figure 5 Decision curve analysis curves of progression-free survival.
A-C: Training set (18 months, 24 months, 36 months); D-F: Validation set (18 months, 24 months, 36 months).
DISCUSSION
This study systematically compared the efficacy and safety of three perioperative regimens, SOX + XDL, P-SOX, and CAPOX, in patients with advanced GC and revealed clinically consistent signals linking short-term treatment responses with long-term survival outcomes. With respect to short-term tumor response, although no statistically significant differences were observed among the three groups in terms of objective response rate based on RECIST 1.1 criteria, the SOX + XDL regimen demonstrated more favorable pathological response characteristics according to TRG assessment, characterized by a higher proportion of TRG 0-2 and a lower proportion of TRG 3. These findings suggest that the immunotherapy-combined regimen is more likely to induce deeper tumor regression. Notably, the radiographic response does not always fully capture the true depth of the histopathological tumor response. In the context of perioperative treatment for GC, pathological response has been widely recognized as an important surrogate endpoint for treatment efficacy and long-term prognosis. The higher TRG in the immunochemotherapy group may reflect enhanced tumor immunogenicity and increased tumor-infiltrating effector T cells, which have been shown to be associated with a positive response to PD-1/PD-L1 blockade and improved survival in patients with GC. A multidimensional tumor-infiltrating immune cell signature was predictive of both immunotherapy response and longer PFS in patients with advanced GC, supporting a mechanistic link between immune cell infiltration and durable clinical benefit[14]. Previous perioperative and neoadjuvant studies have demonstrated that TRG more directly reflects the histopathological effectiveness of treatment and is closely associated with long-term survival outcomes. Multiple cohort studies and systematic analyses have shown that patients who achieve better TRG (or deeper pathological response) are more likely to undergo R0 resection, experience lower recurrence rates, and achieve prolonged OS. In contrast, the radiological response assessed by the RECIST criteria may be influenced by measurement timing, treatment-induced fibrosis or necrosis, and intratumoral heterogeneity and has not consistently been validated as a reliable surrogate endpoint for survival in all settings. Therefore, the results of the present study revealed that the differences in the objective response rate did not reach statistical significance, but the improvement in the TRG and favorable survival trends were clearly biologically and clinically plausible. These findings further support the role of the TRG as an important complementary indicator for evaluating perioperative treatment efficacy and prognostic stratification[15].
With respect to long-term outcomes, survival analyses in the present study demonstrated significant differences in both OS and PFS among the different perioperative treatment regimens. Kaplan-Meier curves indicated that compared with the P-SOX and CAPOX regimens, the SOX + XDL regimen consistently achieved superior OS and PFS throughout the follow-up period, with statistically significant between-group differences. Notably, the survival curves across the three groups were relatively close during the early follow-up phase; however, as the follow-up duration increased, the OS and PFS curves of the SOX + XDL group gradually diverged from those of the comparator groups, with more pronounced separation observed during the mid-to-late follow-up period. These findings suggest that the survival benefit conferred by immunotherapy is somewhat time dependent. These observations are highly consistent with robust evidence from recent landmark clinical trials of immunotherapy in GC. In the first-line treatment of advanced or unresectable GC, the CheckMate 649 trial demonstrated that compared with chemotherapy alone, nivolumab combined with platinum-based chemotherapy significantly improved OS and PFS, with sustained survival advantages during later follow-up, reflecting a characteristic long-term tail effect of immunotherapy[16,17]. Similarly, the ORIENT-16 trial confirmed that sintilimab combined with chemotherapy significantly prolonged OS in a Chinese population and maintained consistent survival benefits across different PD-L1 CPS subgroups. Furthermore, the KEYNOTE-859 study extended these findings to a global, multicenter population, validating the long-term survival advantage of PD-1 inhibitor–based combination therapy in patients with HER2-negative GC[18,19]. Collectively, these studies suggest that the addition of immunotherapy to chemotherapy can convert short-term tumor responses into more durable disease control in a subset of patients. In contrast, the ability of intensified chemotherapy alone to improve long-term survival remains limited. Previous studies have shown that although the introduction of agents such as albumin-bound paclitaxel may enhance short-term response rates to some extent, its impact on OS has been inconsistent and unstable[20]. The ATTRACTION-4 trial conducted in an Asian population demonstrated that nivolumab combined with an oxaliplatin-based regimen significantly improved PFS but failed to achieve a statistically significant OS benefit, highlighting that improvements in PFS do not necessarily translate into sustained OS gains in the absence of effective immune modulation[21]. This phenomenon has also been observed in multiple perioperative and real-world chemotherapy studies, underscoring the inherent limitations of chemotherapy intensification alone in overcoming the long-term risk of recurrence and mortality. Moreover, classic perioperative chemotherapy trials, including the MAGIC, FLOT4, and CLASSIC studies, have shown that systemic therapy can improve long-term outcomes in patients with resectable GC; however, the magnitude of survival benefit remains modest, and disease recurrence continues to be a major determinant of OS. Against this backdrop, the early integration of immunotherapy in combination with platinum-based doublet chemotherapy holds promise for further increasing the long-term survival plateau beyond that achievable with chemotherapy alone. In a real-world perioperative treatment setting, the present study revealed concordant advantages of SOX + XDL for both OS and PFS, providing pragmatic evidence to support the application of immunotherapy-combined chemotherapy strategies in the comprehensive management of advanced GC.
With respect to safety, all three regimens demonstrated generally good tolerability. No intolerable adverse events were observed in patients treated with SOX + XDL. Across all treatment groups, the most common adverse events were gastrointestinal reactions (such as nausea and decreased appetite), peripheral sensory neuropathy, and myelosuppression, the majority of which were grades 1-2 in severity. These findings are consistent with evidence from multiple phase III trials evaluating first-line immunotherapy combined with chemotherapy. In the CheckMate 649 trial, the most frequently reported treatment-related adverse events in the nivolumab plus chemotherapy group remained chemotherapy-associated toxicities, including nausea, diarrhea, and peripheral neuropathy. Similarly, in the ORIENT-16 and ATTRACTION-4 studies, grade ≥ 3 treatment-related adverse events were predominantly hematologic toxicities, such as thrombocytopenia, neutropenia, and anemia, indicating that the burden of severe toxicity was largely driven by the chemotherapy backbone. The KEYNOTE-859 trial further demonstrated that although the incidence of severe treatment-related adverse events increased slightly with the addition of pembrolizumab, the overall safety profile remained manageable and did not substantially alter the established toxicity framework of platinum-based chemotherapy[22-25]. These findings are in line with real-world evidence reported by Kutlu et al[26], who reported that first-line nivolumab plus chemotherapy was generally well tolerated, with most adverse events being grades 1-2 and no unexpected safety signals observed. In the present study, a relatively higher incidence of myelosuppression was observed in the P-SOX group. Although the difference did not reach statistical significance, the trend toward increased cumulative toxicity associated with the three-drug regimen warrants attention, particularly in the context of perioperative treatment, where tolerability and treatment completion are critical considerations.
Several limitations of the present study should be acknowledged. First, immune checkpoint inhibitors may be associated with rare but potentially severe immune-related adverse events, such as grade III-IV pneumonitis or colitis, with reported incidence rates of approximately 5%-10%[27]. Immune-related adverse events were not systematically captured in the present analysis; nevertheless, vigilant monitoring and prompt management remain essential in routine clinical practice when immunotherapy is administered. Second, heterogeneity related to geographic region, patient population, and data processing methods may affect the stability and generalizability of prognostic outcomes. Previous studies have suggested that the prognosis of GC may also be influenced by multiple factors, including age, tumor location, ASA score, abnormal body mass index, the number of dissected lymph nodes, chemotherapy cycles, and postoperative complications[28-34]. These variables could not be comprehensively evaluated in the current study and may have contributed to residual confounding. Third, owing to the retrospective nature of this study, incomplete clinical records resulted in the exclusion of some patients. Fourth, Absence of biomarker data, key predictive markers for immunotherapy response, including PD-L1 CPS and microsatellite instability status, were not assessed. Consequently, the study cannot determine whether observed treatment benefits are enriched in specific molecular subgroups, limiting the precision and generalizability of conclusions regarding immunotherapy utility. Fifth, non-randomized treatment assignment, given the observational nature of the study, regimen selection was influenced by clinical judgment, patient comorbidities, socioeconomic factors, and institutional preferences. Although baseline characteristics were balanced via statistical adjustment, residual confounding remains plausible and may affect causal inference. In addition, the exact proportion of patients who proceeded to surgical resection after perioperative chemotherapy could not be accurately determined. Consequently, only patients who completed both preoperative and postoperative chemotherapy in combination with curative surgery were included, which may have introduced selection bias and limited the representativeness of the study population. Finally, the relatively small sample size and single-center design may reduce the statistical power and external validity of the findings. Future prospective, multicenter studies with larger sample sizes, ideally phase III randomized controlled trials, are warranted to further validate these results.
CONCLUSION
In this real-world cohort of patients with advanced GC who completed perioperative treatment and achieved R0 resection, SOX + XDL was associated with improved pathological response and favorable survival outcomes compared with P-SOX and CAPOX. The observed time-dependent survival advantage may be partly explained by enhanced tumor regression and immune-mediated effects.
Yamada Y, Higuchi K, Nishikawa K, Gotoh M, Fuse N, Sugimoto N, Nishina T, Amagai K, Chin K, Niwa Y, Tsuji A, Imamura H, Tsuda M, Yasui H, Fujii H, Yamaguchi K, Yasui H, Hironaka S, Shimada K, Miwa H, Hamada C, Hyodo I. Phase III study comparing oxaliplatin plus S-1 with cisplatin plus S-1 in chemotherapy-naïve patients with advanced gastric cancer.Ann Oncol. 2015;26:141-148.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 453][Cited by in RCA: 426][Article Influence: 38.7][Reference Citation Analysis (6)]
Shah MA, Shitara K, Ajani JA, Bang YJ, Enzinger P, Ilson D, Lordick F, Van Cutsem E, Gallego Plazas J, Huang J, Shen L, Oh SC, Sunpaweravong P, Soo Hoo HF, Turk HM, Oh M, Park JW, Moran D, Bhattacharya P, Arozullah A, Xu RH. Zolbetuximab plus CAPOX in CLDN18.2-positive gastric or gastroesophageal junction adenocarcinoma: the randomized, phase 3 GLOW trial.Nat Med. 2023;29:2133-2141.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 485][Cited by in RCA: 462][Article Influence: 154.0][Reference Citation Analysis (0)]
Shen L, Zhang Y, Li Z, Zhang X, Gao X, Liu B, Wang Y, Ba Y, Li N, Zhang R, Zhang J, Chen Y, Chen J, Huang M, Fu Y, Liu M, Liu Z, Zhao J, Li W, Wei J, Li C, Xu N, Guo Z, Cao B, Liu L, Nie P, Wan L, Sheng L, Liu Z, He Y, Gu K, Wu G, Wang W, Zhang F, Qiu W, Guo J, Ying J, Pan H, Xu H, Yuan Y, Bai Y, Wang Z, Xu J, Zhao X, Liu H, Zhang X, Dai W, Xu H, Liu M, Xie L, Tang Y, Jin J, Qu X, Fang X, Huang M, Chen H, Zheng Z, Wang Y, Wang D, Li X, Yu G, Liu H, Zhou Y, Zhong D, Zeng S, Kang M, Wang M, Gao Y, Li W, Wang Z, Zhang M, Zhang J, Li Q, Sun S, Zang A, Lin L, Xie M, Zhuang Z, Zhang T, Yao Z, Lu D, Liu W, Hu M, Wang ZM, Li B, Xia M, Zhang J, Ying X, Pardoll DM, Ji J. First-line cadonilimab plus chemotherapy in HER2-negative advanced gastric or gastroesophageal junction adenocarcinoma: a randomized, double-blind, phase 3 trial.Nat Med. 2025;31:1163-1170.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 2][Cited by in RCA: 65][Article Influence: 65.0][Reference Citation Analysis (0)]
Wang FH, Zhang XT, Tang L, Wu Q, Cai MY, Li YF, Qu XJ, Qiu H, Zhang YJ, Ying JE, Zhang J, Sun LY, Lin RB, Wang C, Liu H, Qiu MZ, Guan WL, Rao SX, Ji JF, Xin Y, Sheng WQ, Xu HM, Zhou ZW, Zhou AP, Jin J, Yuan XL, Bi F, Liu TS, Liang H, Zhang YQ, Li GX, Liang J, Liu BR, Shen L, Li J, Xu RH. 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.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 270][Cited by in RCA: 270][Article Influence: 135.0][Reference Citation Analysis (2)]
Bang YJ, Van Cutsem E, Fuchs CS, Ohtsu A, Tabernero J, Ilson DH, Hyung WJ, Strong VE, Goetze TO, Yoshikawa T, Tang LH, Hwang PMT, Webb N, Adelberg D, Shitara K. KEYNOTE-585: Phase III study of perioperative chemotherapy with or without pembrolizumab for gastric cancer.Future Oncol. 2019;15:943-952.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 144][Cited by in RCA: 133][Article Influence: 19.0][Reference Citation Analysis (2)]
Martin-Romano P, Sola JJ, Diaz-Gonzalez JA, Chopitea A, Iragorri Y, Martínez-Regueira F, Ponz-Sarvise M, Arbea L, Subtil JC, Cano D, Ceniceros L, Legaspi J, Hernandez JL, Rodríguez J. Role of histological regression grade after two neoadjuvant approaches with or without radiotherapy in locally advanced gastric cancer.Br J Cancer. 2016;115:655-663.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 39][Cited by in RCA: 39][Article Influence: 3.9][Reference Citation Analysis (0)]
Moehler M, Xiao H, Blum SI, Elimova E, Cella D, Shitara K, Ajani JA, Janjigian YY, Garrido M, Shen L, Yamaguchi K, Liu T, Schenker M, Kowalyszyn R, Bragagnoli AC, Bruges R, Montesarchio V, Pazo-Cid R, Hunter S, Davenport E, Wang J, Kondo K, Li M, Wyrwicz L. Health-Related Quality of Life With Nivolumab Plus Chemotherapy Versus Chemotherapy in Patients With Advanced Gastric/Gastroesophageal Junction Cancer or Esophageal Adenocarcinoma From CheckMate 649.J Clin Oncol. 2023;41:5388-5399.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 19][Cited by in RCA: 36][Article Influence: 12.0][Reference Citation Analysis (0)]
Janjigian YY, Ajani JA, Moehler M, Shen L, Garrido M, Gallardo C, Wyrwicz L, Yamaguchi K, Cleary JM, Elimova E, Karamouzis M, Bruges R, Skoczylas T, Bragagnoli A, Liu T, Tehfe M, Zander T, Kowalyszyn R, Pazo-Cid R, Schenker M, Feeny K, Wang R, Lei M, Chen C, Nathani R, Shitara K. First-Line Nivolumab Plus Chemotherapy for Advanced Gastric, Gastroesophageal Junction, and Esophageal Adenocarcinoma: 3-Year Follow-Up of the Phase III CheckMate 649 Trial.J Clin Oncol. 2024;42:2012-2020.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 204][Cited by in RCA: 185][Article Influence: 92.5][Reference Citation Analysis (6)]
Xu J, Jiang H, Pan Y, Gu K, Cang S, Han L, Shu Y, Li J, Zhao J, Pan H, Luo S, Qin Y, Guo Q, Bai Y, Ling Y, Yang J, Yan Z, Yang L, Tang Y, He Y, Zhang L, Liang X, Niu Z, Zhang J, Mao Y, Guo Y, Peng B, Li Z, Liu Y, Wang Y, Zhou H; ORIENT-16 Investigators. Sintilimab Plus Chemotherapy for Unresectable Gastric or Gastroesophageal Junction Cancer: The ORIENT-16 Randomized Clinical Trial.JAMA. 2023;330:2064-2074.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 316][Cited by in RCA: 319][Article Influence: 106.3][Reference Citation Analysis (1)]
Kang YK, Chen LT, Ryu MH, Oh DY, Oh SC, Chung HC, Lee KW, Omori T, Shitara K, Sakuramoto S, Chung IJ, Yamaguchi K, Kato K, Sym SJ, Kadowaki S, Tsuji K, Chen JS, Bai LY, Oh SY, Choda Y, Yasui H, Takeuchi K, Hirashima Y, Hagihara S, Boku N. Nivolumab plus chemotherapy versus placebo plus chemotherapy in patients with HER2-negative, untreated, unresectable advanced or recurrent gastric or gastro-oesophageal junction cancer (ATTRACTION-4): a randomised, multicentre, double-blind, placebo-controlled, phase 3 trial.Lancet Oncol. 2022;23:234-247.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 671][Cited by in RCA: 624][Article Influence: 156.0][Reference Citation Analysis (5)]
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: 2323][Article Influence: 464.6][Reference Citation Analysis (9)]
Qin S, Bai Y, Li J, Pan H, Luo S, Qu Y, Ye F, Yang L, Liu T, Li W, Chen X, Yang J, Ying J, Lin X, Zhao L, Liang X, Mao Y, Guo R, Zuo Y, Bordia S, Li S. First-Line Pembrolizumab Plus Chemotherapy for HER2-Negative Advanced Gastric Cancer: China Subgroup Analysis of the Randomized Phase 3 KEYNOTE-859 Study.Adv Ther. 2025;42:1892-1906.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in RCA: 9][Reference Citation Analysis (0)]
Janjigian YY, Shitara K, Moehler M, Garrido M, Salman P, Wyrwicz L, Yamaguchi K, Skoczylas T, Bragagnoli AC, 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. A plain language summary of the CheckMate 649 study: nivolumab in combination with chemotherapy compared to chemotherapy alone for untreated advanced or metastatic cancer of the stomach or esophagus.Future Oncol. 2023;19:739-752.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 1][Cited by in RCA: 4][Article Influence: 1.3][Reference Citation Analysis (0)]
Kutlu Y, Dae SA, Yilmaz F, Erdem D, Sendur MAN, Akbas S, Senocak Tasci E, Bas O, Dane F, Sakin A, Kaya AO, Aykan MB, Ergun Y, Biter S, Disel U, Korkmaz M, Selcukbiricik F, Kose F, Olmez OF, Bilici A, Demir G, Yalcin S. Real-World Efficacy and Safety of First-Line Nivolumab Plus Chemotherapy in Patients with Advanced Gastric, Gastroesophageal Junction, and Esophageal Adenocarcinoma: A Nationwide Observational Turkish Oncology Group (TOG) Study.Cancers (Basel). 2024;16:2251.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 1][Cited by in RCA: 5][Article Influence: 2.5][Reference Citation Analysis (0)]
