Construction of competing-risk nomograms and identification of optimal candidates for aggressive therapy in gastric cancer with peritoneal metastasis: a population-based study.

Background: The optimal treatment strategies for gastric cancer with peritoneal metastasis (GCPM) are extensively debated, especially regarding the appropriateness of aggressive local treatments. This study aimed to construct robust competing-risk nomograms for prognostic prediction and to establish a novel risk-stratification system to facilitate individualized therapeutic decision-making. Methods: Patients diagnosed with GCPM between 2010 and 2015 were identified from the Surveillance, Epidemiology, and End Results (SEER) database and randomly divided into training and validation cohorts at a 7:3 ratio. Propensity score matching (PSM) was applied to evaluate the survival impact of different treatment strategies. Independent prognostic factors in the training cohort were selected using a combination of least absolute shrinkage and selection operator (LASSO) regression and stepwise Akaike information criterion (AIC), and these variables were subsequently used to construct Cox [for overall survival (OS)] and Fine-Gray [for cancer-specific survival (CSS)] nomograms. Model discrimination, calibration, predictive accuracy, and clinical utility were assessed using the concordance index (C-index), receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). Risk stratification was further performed based on the nomogram-derived scores, and treatment benefits were analyzed across different risk groups. Results: A total of 4,280 patients were included, comprising 2,996 in the training cohort and 1,284 in the validation cohort. PSM analysis demonstrated that, among unstratified patients, aggressive therapy significantly improved survival compared with chemotherapy-based treatment (P<0.001). Twelve independent prognostic factors, including treatment strategy, tumor grade, and metastatic burden, were identified for the construction of nomograms. The OS and CSS nomograms demonstrated favorable discrimination [C-index: 0.68-0.71; 2-year area under the curve (AUC): 0.792; 95% confidence interval (CI): 0.769-0.815] and calibration in both cohorts, while DCA suggested a potential for greater clinical net benefit than the American Joint Committee on Cancer (AJCC) staging system across a range of threshold probabilities. Risk stratification utilizing nomogram scores effectively distinguished subgroups with different prognostic levels. Treatment-benefit analyses revealed that aggressive therapy was associated with significantly reduced the risk of death in the low-risk group [hazard ratio (HR) =0.6, P<0.001], whereas it was associated with increased mortality risk in the high-risk group (characterized by advanced age and extensive tumor burden; HR >1, P<0.001). Conclusions: This study developed and validated competing-risk nomograms for GCPM patients and proposed a new risk-stratification system. This system supports the concept of risk-adapted therapeutic strategies, suggesting that low-risk patients may be more likely to derive survival benefits from aggressive therapy, while high-risk patients are better suited for systemic chemotherapy or palliative care to avoid ineffective and potentially harmful overtreatment.