FLOT versus SOX as immunochemotherapy backbones for neoadjuvant or conversion therapy in gastric cancer: efficacy, toxicity, and postoperative safety in a real-world cohort.

Objective: To compare FLOT and SOX as chemotherapy backbones for neoadjuvant or conversion immunochemotherapy in gastric cancer, with emphasis on pathological efficacy, treatment-related toxicity, treatment feasibility, postoperative safety, and exploratory survival outcomes. Methods: This retrospective real-world cohort study included 148 patients with gastric cancer who received preoperative immune checkpoint inhibitor-based therapy, with or without targeted therapy, followed by gastrectomy with D2 lymphadenectomy from 2021 to 2025. Patients were grouped according to chemotherapy backbone: FLOT-based immunochemotherapy (n = 75) or SOX-based immunochemotherapy (n = 73). The primary endpoint was grade 3 or higher treatment-related adverse events. Secondary endpoints included pathological complete response, major pathological response, ypN0 status, R0 resection, adverse event-related treatment modification, postoperative complications, disease-free survival, and overall survival. Exploratory survival analyses included Kaplan-Meier estimation and Firth-penalized Cox proportional hazards regression to account for the limited number of survival events. All subgroup analyses were exploratory and hypothesis-generating. Results: FLOT and SOX achieved comparable pathological efficacy. Major pathological response occurred in 42.7% of patients in the FLOT group and 45.2% in the SOX group, while pathological complete response occurred in 18.7% and 27.4%, respectively. ypN0 status, clinical-to-pathological downstaging, and R0 resection rates were also similar. Grade 3 or higher adverse events were numerically more frequent with FLOT than SOX (46.7% vs. 32.9%), as were adverse event-related dose reduction or treatment interruption (36.0% vs. 21.9%). However, postoperative complications were comparable between groups (21.3% vs. 23.3%). Sensitivity analyses excluding targeted therapy showed consistent findings. Exploratory survival analyses were limited by 32 DFS events and 16 deaths and by markedly shorter follow-up in the SOX group. Although the unadjusted hazard ratios numerically favored FLOT for both DFS and OS, multivariable Firth-penalized Cox models substantially attenuated these estimates (DFS adjusted HR, 1.08; 95% CI, 0.50-2.34; OS adjusted HR, 1.56; 95% CI, 0.53-4.60). Accordingly, no definitive conclusions regarding survival superiority, inferiority, or equivalence could be drawn. Conclusion: In this cohort, FLOT and SOX showed broadly comparable pathological efficacy and postoperative complication rates, whereas FLOT was associated with a numerically greater preoperative toxicity and treatment-modification burden. SOX may therefore represent a balanced option in selected patients, but these findings require prospective validation.