Predicting the grade of dysplasia of pancreatic cystic neoplasms using cyst fluid DNA methylation markers

Purpose: Pancreatic cysts are common and pose diagnostic and management challenges. Pancreatic cyst fluid markers have the potential to aid in the management of cysts with concerning imaging findings. Our aim was to evaluate cyst fluid methylated DNA markers for their accuracy for predicting the histologic grade of neoplastic pancreatic cysts. Experimental Design: Pancreatic cyst fluid samples from 183 patients (29 discovery and 154 validation) aspirated after surgical resection were analyzed for methylated DNA at selected genes (SOX17, BNIP3, FOXE1, PTCHD2, SLIT2, EYA4, and SFRP1) using methylation-specific droplet-digital PCR (dd-QMSP). Methylated DNA levels were evaluated for their accuracy at predicting the grade of dysplasia of the pancreatic cyst. Results: All six markers evaluated in the validation set could accurately distinguish high-risk cystic neoplasms (with high-grade dysplasia and/or associated invasive cancer) from low-risk cysts (lower grades of dysplasia) with accuracies from 79.8% to 83.6%. Methylated SOX17 had the highest overall accuracy as a single marker (sensitivity, 78.4%; specificity, 85.6%; accuracy 83.6%, cutoff; 25 methylated DNA molecules/μL cyst fluid). The best four-gene combination had 84.3% sensitivity, 89.4% specificity, and 88.0% accuracy at distinguishing cysts with high-grade dysplasia and/or invasive cancer from those without. All six markers were independent predictors of having invasive cancer/high-grade dysplasia after adjusting for clinical/imaging factors known to be associated with grade of dysplasia. The combination of methylated SOX17 with cytology better predicted neoplastic grade than cytology alone. Conclusions: A panel of methylated gene markers quantified by dd-QMSP can be used to predict the grade of dysplasia of pancreatic cysts.