Clinicopathological Factors on Survival after Conversion Surgery for Unresectable Locally Advanced Pancreatic Cancer: A Nationwide Study by the Japanese Society of Hepato-Biliary-Pancreatic Surgery.

Yasuda, Satoshi; Satoi, Sohei; Takami, Hideki; Hirano, Satoshi; Akita, Hirofumi; Takahashi, Yu; Yoshida, Yuta; Unno, Michiaki et al. · Ann Surg · 2026

retrospective_cohort · Level III

Where this comes from

Abstract

To identify prognostic factors, including preoperative treatment duration, among patients who underwent conversion surgery (CS) for unresectable locally advanced pancreatic cancer (UR-LAPC). While CS has been increasingly adopted for UR-LAPC, optimal perioperative strategies remain controversial. This multicenter study included 465 UR-LAPC patients who underwent CS following preoperative chemotherapy with FOLFIRINOX (FFX) or gemcitabine plus nab-paclitaxel (GnP) from 2015 to 2020 at 84 Japanese institutions. Median overall survival (OS) from treatment initiation was 43.8 months with a 5-year survival rate of 37.2%. A prognostic cutoff for preoperative treatment duration was identified at 6.1 months using maximally selected rank statistics. Patients receiving >6 months of preoperative treatment (n=350) demonstrated significantly better OS (50.4 vs. 29.7 mo) and recurrence-free survival (RFS) (15.6 vs. 9.1 mo) compared with those receiving ≤6 months (n=115, both P<0.001). Multivariate analysis identified four independent preoperative prognostic factors: treatment duration >6 months, FFX-based regimens, normal tumor markers (CA19-9 and CEA), and a prognostic nutritional index ≥45 before CS. These four preoperative factors enabled clear prognostic stratification: patients with ≥3 factors showed significantly improved survival compared with those with ≤2 factors (HR 0.44, P<0.0001; 5-year OS: 59.8% vs. 26.3%). The combination of four preoperative prognostic factors may enable risk stratification among patients undergoing CS for UR-LAPC. These findings may help inform treatment sequencing and patient selection, although external validation is needed to confirm their generalizability.