ABSTRACT Pancreatic cystic lesions (PCLs) are the only radiographically identifiable precursor to deadly pancreatic carcinoma and they are detected in a million people annually in the U.S. Most patients with PCLs have “indeterminate” cysts in which both the type and histologic grade are clinically and radiographically uncertain. Currently used cyst fluid diagnostics suffer from poor accuracy and they typically require large volumes of fluid that are unavailable in ~50% of cases, leading to both missed cancers and unnecessary major surgery. In this proposal, we leverage our novel multiplex substrate profiling by mass spectrometry method combined with proteomics to refine target proteins and then validate our two-tier classifier for the identification of mucinous PCLs with high grade dysplasia or invasive carcinoma, together considered advanced neoplasia (AN). The final clinical assay requires only 5 µL of fluid. A micro-volumetric assay is expected to improve the diagnostic yield of cyst fluid testing from ~50% to >90% for patients who undergo routine invasive testing. The Tier-1 classifier identifies mucinous PCLs, which carry variable risk of malignant transformation, thereby potentially removing ~25% of patients with non-mucinous PCLs from the need for further surveillance or surgery. We will validate the Tier-1 classifier with PCL samples from the Early Detection Research Network (EDRN), a histologically diverse, well- annotated biorepository + untested samples from UCSF and UCLA. We will further test and refine our Tier-2 classifier for the identification of AN using untested banked mucinous PCLs from UCSF, and then validate it using external samples from the EDRN and prospectively collected samples from UCSF and UCLA to prioritize detection of AN while minimizing the burdens of overtreatment. We will also use our radiology search engine, mPower, to improve capture of “missed” patients with incidental clinically relevant PCLs, and we will thereby expand the racial and ethnic diversity of our Panc Cyst Registry. Our registry is a prospective longitudinal research resource that uses patient-directed input and includes demographic and lifestyle information. Together these novel tools are expected to make a significant impact on the management of “indeterminate” PCLs and improve the detection of early-stage pancreatic cancer, while reducing the burdens of overdiagnosis and overtreatment.

PROJECT NARRATIVE Pancreatic cystic lesions (PCLs) are incidentally identified on abdominal imaging studies in more than a million patients annually in the U.S. and they represent an opportunity for early detection of pancreatic adenocarcinoma. Most patients with PCLs have “indeterminate” cysts in which both the type and histologic grade are clinically and radiographically uncertain, and current diagnostic tests are inadequate and require large cyst fluid volumes that are often unavailable. We propose to evaluate our novel micro-volumetric two-tier classifier that can improve diagnostic accuracy of current cyst fluid tests, as well as to use novel tools including mPower, a radiology report search engine, to improve capture of patients with “missed” PCLs and, thereby, expand racial and ethnic diversity in our electronic Panc Cyst Registry.