Multianalyte blood-based risk stratification of incidental pancreas lesions

Abstract

Abstract Purpose Incidental detection of pancreas lesions (IPLs) is common and creates an opportunity to intercept pancreatic ductal adenocarcinoma (PDAC). However, identifying patients at risk for progression who can benefit from surgical resection remains challenging. The current role of blood tests for risk stratification in patients with IPLs is limited. Methods We evaluated the performance of circulating glycoproteins (CA19-9, CEA and CA125), plasma DNA fragmentation analysis, and their combination, in 99 asymptomatic patients with IPLs, for detection of advanced pathology (high-grade dysplasia or invasive carcinoma). Plasma DNA fragmentation was analyzed using a machine learning model, adapted to detect PDAC in 242 patients with cancer and 300 healthy individuals. Results During 18.8 months of median follow-up by a multidisciplinary clinical team, 11 of 99 patients with IPLs were diagnosed with advanced pathology. We observed area under the receiver operating characteristic curves (AUROCs) of 0.78, 0.63, 0.54 and 0.74 using CA19-9, CEA, CA125 and plasma DNA fragmentation, respectively. Combined analysis of CA19-9, CA125 and plasma DNA fragmentation showed an AUROC of 0.93, with 91% sensitivity and 53% positive predictive value (PPV) at 90% specificity. In a subset of 25 patients with histologically confirmed diagnoses, AUROC improved to 0.96 and PPV improved to 91%. In one patient with an equivocal initial endoscopy, multi-analyte blood analysis predicted cancer 3 months before diagnosis of stage IA cancer. Conclusion Our results demonstrate a multianalyte blood test combining glycoprotein biomarkers with plasma DNA fragmentation could complement current clinical workflows for cancer detection in patients with IPLs.

Ähnliche Arbeiten