Pancreatic ductal adenocarcinoma (PDAC) is an emerging health issue, being estimated to become the 2nd leading cause of death by 2030. The only chance for long-term cure is surgery. Patients with borderline or locally advanced disease (clinical nodal disease, high CA19.9 or vascular involvement) are proposed for neaodjuvant treatment while clinically resectable ones, according to the current guidelines, are sent for upfront surgery. Unfortunately up to 40% of patients scheduled for curative resection will be found unresectable at laparotomy: their identification would allow a better resources allocation. Preoperative predictors of markers such as CA19.9, widely used in prognosis, has several limitations, being undetectable in 10% of patients. Even imaging is unreliable in predicting resectability, having shown to systematically under-stage patients. Recently the correlation of image features from radiographic images, also called Radiomics, and survival outcomes has been validated in lung cancer as well as in colorectal cancer and few promising papers are coming out even in pancreatic cancer. A group from MSKCC, NY, has developed a model based on image features, CA19.9 and a pathologic score called Brennan score to predict survival in resected PDAC patients. This valuable model has the defect of giving an information based on postoperative features, hence too late. Circulating-free DNA has been shown to be significantly elevated in PDAC and liquid biopsy in such patients is an open and expanding research field. Our goal is to provide a new non-invasive diagnostic tool in the treatment decisional algorithm by combining the prognostic value of CA19.9, radiomics and cfDNA to predict resectability and survival in PDAC patients selected for upfront surgery.
