Abstract
Pancreatic ductal adenocarcinoma (PDAC), a "silent killer" with elusive early symptoms and poor prognosis, sees nearly half of patients experience recurrence within a year post-curative-intent surgery. Very early recurrence (VER), defined as recurrence within 12 weeks postoperatively and first termed "biological R2 resection" by Belfiori et al, remains a clinical puzzle. Martlı et al's recent retrospective cohort study offers crucial insights into this understudied issue, identifies predictive factors that challenge long-held beliefs, and calls for a rethink of risk stratification and postoperative management for PDAC patients. Martlı et al studied 303 PDAC patients at a high-volume center from 2019 to 2024, with VER affecting 9.24% (28 patients) of the cohort. The study's strength lies in combining traditional statistical analyses and machine learning (random forest modeling) to capture nonlinear relationships between clinicopathological factors and VER risk. Key findings include: (1) Poorly differentiated (G3) tumors are the strongest VER predictor (OR = 2.43, P < 0.001; random forest importance score = 0.35), with 92.85% of VER patients having G3 tumors (vs 45.81% of non-VER patients); (2) Contrary to prior studies, pancreatic head tumors (89.28% of VER patients vs 83.66% of non-VER patients, P = 0.031) were linked to VER; (3) Elevated red cell distribution width is a weaker predictor (random forest importance score = 0.20, P = 0.03 for group difference, P = 0.079 in multivariate analysis); and (4) VER correlates with significantly higher 6-month mortality (32.44% vs 14.77% in non-VER patients, P = 0.032).