Background: The optimal number of retrieved lymph nodes (LNs) in gastric cancer (GC) is still debatable and previous studies proposing new classification alternatives mostly focused on the number of retrieved LNs without proper consideration on the anatomic nodal groups’ location. Here, we assessed the impact of retrieved LNs from different nodal location groups on the survival of GC patients. Methods: Stage I–III gastric cancer patients who had radical gastrectomy were investigated. LN grouping was determined according to the 13th edition of the JCGC. The optimal cut‐off values of retrieved LNs in different LN groups (Group 1 and 2) were calculated, based on which a proposed nodal classification (rN) simultaneously accounting the optimal number and location of retrieved LNs was proposed. The performance of rN was then compared to that of LN ratio, log‐odds of metastatic LNs (LODDs) and the 8th edition of the Union for International Cancer Control/American Joint Committee on Cancer (UICC/AJCC) N classification. Results: The optimal cut‐off values for Group 1 and 2 were 13 and 9, respectively. The 5‐year overall survival (OS) was higher for patients in retrieved Group 1 LNs > 13 (vs. Group 1 LNs ≤ 13, 63.2% vs. 57.9%, P = 0.005) and retrieved Group 2 LNs > 9 (vs. Group 2 LNs ≤ 9, 72.5% vs. 60.7%, P = 0.009). Patients staged as pN0–3b were sub classified using this Group 1 and 2 nodal analogy. The OS of pN0–N2 patients in retrieved Group 1 LNs > 13 or Group 2 LNs > 9 were superior to those in retrieved Group 1 LNs ≤ 13 and Group 2 LNs ≤ 9 (All P < 0.05); except for pN3 patients. The rN classification was formulated and demonstrated better 5‐year OS prognostication performance as compared to the LNR, LODDs, and the 8th UICC/AJCC N staging system. Conclusions: The retrieval of > 13 and > 9 LNs for Group 1 and Group 2, respectively, could represent an alternative lymph node retrieval approach in radical gastrectomy for more precise survival prognostication and minimizing staging migration, especially if > 16 LNs is found to be difficult.

Background. In gastric cancer, various surveillance strategies are suggested in international guidelines. The current study is intended to evaluate the current strategies and provide more personalized proposals for personalized cancer medicine. Materials and Methods. In the aggregate, 9191 patients with gastric cancer after gastrectomy from 1998 to 2009 were selected from the Surveillance, Epidemiology, and End Results database. Disease-specific survival was analyzed by Kaplan-Meier method and the log-rank test. Cox proportional hazards regression analyses were used to confirm the independent prognostic factors. As well, hazard ratio (HR) curves were used to compare the risk of death over time. Conditional survival (CS) was applied to dynamically assess the prognosis after each follow-up. Results. Comparisons from HR curves on different stages showed that earlier stages had distinctly lower HR than advanced stages. The curve of stage IIA was flat and more likely the same as that of stage I while that of stage IIB is like that of stage III with an obvious peak. After estimating CS at intervals of three months, six months, and 12 months in different periods, stages I and IIA had high levels of CS all along, while there were visible differences among CS levels of stages IIB and III. Conclusions. The frequency of follow-up for early stages, like stages I and IIA, could be every six months or longer in the first three years and annually thereafter. And those with unfavorable conditions, such as stages IIB and III, could be followed up much more frequently and sufficiently than usual.

Early detection and precise preoperative staging of early gastric cancer (EGC) are critical. Therefore, this study aims to develop a deep learning model using portal venous phase CT images to accurately distinguish EGC without lymph node metastasis. This study included 3164 patients with gastric cancer (GC) who underwent radical surgery at two medical centers in China from 2006 to 2019. Moreover, 2.5D radiomic data and multi-instance learning (MIL) were novel approaches applied in this study. By basing the selection of features on 2.5D radiomic data and MIL, the ResNet101 model combined with the XGBoost model represented a satisfactory performance for diagnosing pT1N0 GC. Furthermore, the 2.5D MIL-based model demonstrated a markedly superior predictive performance compared to traditional radiomics models and clinical models. We first constructed a deep learning prediction model based on 2.5D radiomics and MIL for effectively diagnosing pT1N0 GC patients, which provides valuable information for the individualized treatment selection.