-
A risk score system including CT features for predicting early recurrence of resectable pancreatic ductal adenocarcinoma after radical resection: a dual-center retrospective study
Yan Deng, Haopeng Yu, Marc Zins, Xueying Zhang, Xiuping Duan, Xiaoming Zhang, Yi Zhang, Dan Cao, Zixing Huang, Bin Song
Eur Radiol. 2025 May 2. doi: 10.1007/s00330-025-11632-y. Online ahead of print.Abstract
Purpose: To develop a score system including CT features for predicting postoperative early (≤ 1 year) recurrence-free survival (RFS) in resectable pancreatic ductal adenocarcinoma (PDAC) patients who underwent radical resection and assess its performance.
Materials and methods: This dual-center, retrospective study included patients with resectable PDAC who underwent radical resection from September 2016 to April 2023. All CT features were independently evaluated by two blinded radiologists. An early recurrence score (ERS) based on CT and clinical features, for predicting early recurrence risk, was developed by Cox regression analysis in the developing cohort, and was validated in the testing and validation cohorts and compared with AJCC TNM staging system.
Results: This study included 210 patients in the development cohort (mean age � standard deviation, 60 � 10 years; 129 men), 92 patients in the testing cohort (60 � 9 years; 60 men), and 31 patients in the validation cohort (62 � 7 years; 20 men). CA19-9 (hazard ratio [HR], 1.57; p = 0.044), perineural invasion on CT (HR, 1.83; p = 0.037), tumor necrosis (HR, 3.20; p < 0.001), and lymph nodes metastasis on CT (HR, 1.84; p = 0.004) formed the ERS. Its AUC of 0.851 and 0.901, superior to AJCC TNM staging (AUC of 0.630 and 0.534) in the testing and validation cohorts, (p < 0.05), respectively. The high-risk patients predicted by ERS had significantly higher postoperative 1-year recurrence rates than their low-risk counterparts in both the testing cohort (81.4% vs 22.5%, p < 0.001) and the validation cohort (81.2% vs 26.7%, p = 0.003).
Conclusion: The ERS noninvasively predicted early recurrence in resectable PDAC, outperforming the AJCC TNM system.