Multiinstitutional Validation Study of Cyst Fluid Protein Biomarkers in Patients With Cystic Lesions of the Pancreas.

ABSTRACT

OBJECTIVE: Prospective evaluation of 2 clinical-molecular models in patients with unknown pathology who underwent endoscopic ultrasound with fine-needle aspiration (EUS-FNA) for a cystic lesion of the pancreas. SUMMARY OF BACKGROUND DATA Preoperative prediction of histologic subtype (mucinous vs nonmucinous) and grade of dysplasia in patients with pancreatic cystic neoplasms is challenging. Our group has previously published 2 clinical-molecular nomograms for intraductal papillary mucinous neoplasms (IPMN) that incorporated both clinical/radiographic features and cyst fluid protein markers (sFASL, CA72-4, MMP9, IL-4). METHODS: This multiinstitutional study enrolled patients who underwent EUS-FNA for a cystic lesion of the pancreas. Treatment recommendations regarding resection were based on standard clinical, radiographic, and endoscopic features. Predicted probabilities of high-risk IPMN (high-grade dysplasia/invasive cancer) were calculated using the previously developed clinical-molecular nomograms. RESULTS: Cyst fluid was obtained from 100 patients who underwent diagnostic EUS-FNA. Within this group there were 35 patients who underwent resection, and 65 were monitored radiographically. Within the group that underwent resection, 26 had low-risk IPMN or benign non-IPMN lesions, and 9 had high-risk IPMN. Within the surveillance group, no patient progressed to resection or developed cancer after a median follow-up of 12months (range 0.5-38). Using the clinical/radiographic nomogram alone, 2 out of 9 patients with high-risk IPMN had a predicted probability >0.5. In the clinical-molecular models, 6 of 9 patients in model 1, and 6 of 9 in model 2, had scores >0.5. CONCLUSIONS: This prospective study of patients with unknown cyst pathology further demonstrates the importance of cyst fluid protein analysis in the preoperative identification of patients with high-risk IPMN. Longer follow-up is necessary to determine if this model will be useful in clinical practice.