Abstract 6641: Identifying candidates for upfront R0 gastrectomy in peritoneal oligometastatic gastric cancer.

Abstract Importance Surgeons frequently encounter gastric cancer with peritoneal oligometastasis in situations where conversion surgery is not feasible. However, objective criteria for selecting candidates for upfront curative- intent R0 gastrectomy remain undefined. Objective To develop and externally validate a machine learning model and a simplified bedside score to stratify patients with gastric cancer and peritoneal oligometastasis undergoing upfront curative-intent R0 gastrectomy. Design, Setting, and Participants This retrospective, multicenter cohort study was conducted from 2014 to 2021 across six high-volume institutions in South Korea. The model was trained using data from four institutions and externally validated using data from two others. A total of 792 patients with gastric cancer and synchronous peritoneal metastasis without other distant metastases were screened. The final cohort included 80 patients with oligometastatic peritoneal carcinomatosis (P1/P2) who underwent R0 gastrectomy. Exposure Upfront curative-intent R0 gastrectomy, defined as gastrectomy with complete peritoneal metastasectomy. Main Outcomes and Measures The primary outcome was overall survival. Model performance was evaluated using the concordance index. An optimal risk score cutoff was determined by ROC analysis in the training set to classify patients into low- and high-risk groups. Secondary outcomes included peritoneal progression-free survival and overall survival after peritoneal progression. Agreement between patient stratification by the simple risk score and the machine learning model was assessed using accuracy and F1-score. Results The machine learning model effectively stratified patients, showing strong performance in the validation cohort (C-index, 0.811). Low-risk patients had significantly longer overall survival (log-rank p 0.001), progression-free survival (log-rank p 0.001), and survival after peritoneal progression (log-rank p 0.001) compared with high-risk patients. Notably, high-risk patients derived no survival benefit from R0 gastrectomy compared with R2 gastrectomy or no gastrectomy, whereas low-risk patients achieved a 5- year survival rate exceeding 30%. The simplified 8-item risk score showed high concordance with model- based stratification (accuracy = 0.9125, F1-score = 0.9114). Conclusions and Relevance This study presents a validated machine learning model and an accompanying simple risk scoring system that stratify patients with peritoneal oligometastatic gastric cancer according to survival benefit from upfront R0 gastrectomy. These tools offer an objective, data-driven framework for intraoperative decision-making, enabling avoidance of futile surgery in high-risk patients and timely R0 gastrectomy in low-risk candidates, thereby facilitating personalized surgical strategies. Citation Format: Hyoung-Il KIM, Seungho Lee. Identifying candidates for upfront R0 gastrectomy in peritoneal oligometastatic gastric cancer abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6641.