Expanding loco-regional nodes harvesting is expected to increase survival. This improvement may be associated to stage migration (SM). However, the great bulk of harvested lymph nodes observed in large dissections is negative. M&M: 830 patients who received R0 gastrectomy for adenocarcinoma were included. pN+ patients with <26 nodes (n = 209) were included for a simulation to "offer 26 nodes" - SM (proportional and exponential based) was simulated and analysed through machine learning algorithms. Overall Survival (OS), in native and simulated stages, were compared. OS of extended lymphadenectomies (pN+, D ≥ 26, n = 273) was compared with the simulated curves. OS of patients in the following dissection intervals of negative nodes were compared: <16 (n = 233), 16-25 (n = 258), ≥26 (n = 339). RESULTS: After simulation to 26 nodes (pN+, D < 26 patients, n = 209), staging was recomputed. OS of native vs simulated early-stages (I-II) and advanced stages (III) were not different (p > 0.05). OS of patients with lymphadenectomy (≥26) was better than simulated for early and advanced stages (p = 0.008; p = 0.005). OS of patients included in distinct intervals of negative lymph nodes were different (p < 0.001). These intervals were an independent prognostic factor (multivariate analysis). CONCLUSIONS: The influence of Stage Migration was null in this set of simulations and Will Rogers phenomenon was not observed. Extended dissection performed better in OS. But the influence of the number of negative nodes, even in large dissections, was highlighted. By emphasizing the role of negative nodes, we aim to facilitate more informed decision-making in management of gastric cancer patients, ultimately leading to improved treatment outcomes and patient care.
Keywords: Gastric cancer; Lymphadenectomy; Machine learning; Negative lymph node count; Overall survival; Simulation; Stage migration.
© 2024 Published by Elsevier Ltd.