The current TNM staging system is not sufficient for prediction of prognosis and cannot precisely identify the patients who are in greater need of adjuvant therapy in pancreatic ductal adenocarcinoma (PDAC). Tumor mutation and copy number variation (CNV) markers may have a higher predictive value. In this study, whole exosome sequencing was performed for patients with stage I-II PDAC undergoing R0 resection. The investigators aimed to identify genes with discrepant statuses of mutations or CNVs between patients with and without relapse within 1 year after R0 resection, and then to construct a support vector machine (SVM)-based prognostic classifier (the SVM signature) for PDAC using machine learning; the investigators then aimed to further validate the SVM signature in an independent cohort.

Inclusion Criteria:

• Availability of hematoxylin and eosin slides with invasive tumor components
• Availability of clinicopathologic characteristics and follow-up data
• No previous history of cancer

Exclusion Criteria:

• No formalin-fixed, paraffin-embedded (FFPE) tumor sample of primary tumor
• Receipt of any neoadjuvant and/or adjuvant cancer-directed therapy
• Survival time <3 months after resection

The investigators enrolled two independent cohorts of consecutive patients with incident, primary, microscopically-confirmed stage I-II PDAC undergoing radical R0 resection only. For the training cohort, data were retrieved for 30 patients undergoing resection between March 2015 and December 2016 at Chinese PLA General Hospital in Beijing, China. Patients who did not relapse or die within 1 year after resection had follow-up of at least 1 year. The investigators also included 40 patients from The Cancer Genome Atlas (TCGA) database (https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga) as the validation cohort, with the same criteria as above; they were diagnosed between January 2010 and December 2013.

Not decided yet