Tree-based model for diffuse-type gastric cancer prognostication: a population study based on the Surveillance, Epidemiology, and End Results database.

Background: Diffuse-type gastric cancer (DGC), a subtype of the Lauren classification, is characterized by poor differentiation and a poor prognosis, and effective therapeutic options remain limited. This study aimed to explore the prognostic risk stratification of DGC and to evaluate the effects of various clinical factors on prognosis. Methods: The data of 1,310 DGC patients were collected from the Surveillance, Epidemiology, and End Results (SEER) database from 2004 to 2015. Cox proportional hazards regression, survival trees, and random survival forests were used for multivariable analyses. The relative importance of different clinical variables was determined by the random survival forest analysis for overall survival (OS) and cancer-specific survival (CSS) in DGC. Results: The 1,310 DGC patients had a median survival time of 23 months. Most of the patients were White and presented with poorly differentiated tumors. Three distinct subgroups and four subgroups were identified by survival tree analyses based on OS and CSS, respectively. For OS, the 5-year rates of the three groups were 79.4%, 37.2%, and 11%, while the 10-year rates were 71.4%, 24.3%, and 7%, respectively. For CSS, the 5-year rates of the four groups was 78.7%, 39.3%, 21.4%, and 7.8%, and the 10-year rates were 70.7%, 30.1%, 16.3%, and 6%, respectively. The random survival forest results highlighted the importance of stage, surgery, age, tumor size, and receipt of chemotherapy in determining both OS and CSS. Conclusions: This study identified different prognostic groups and established a risk stratification framework for DGC. Tumor stage and surgery significantly influence risk stratification. The study findings may inform long-term surveillance strategies and guide treatment decision-making for patients with DGC.