Prediction of malignancy in cystic neoplasms of the pancreas: a population-based cohort study.

OBJECTIVES: Pancreatic cystic neoplasms (PCNs) are being detected with increased frequency. The aims of this study were to determine the incidence of malignancy and develop an imaging-based system for prediction of malignancy in PCN.

METHODS: We conducted a retrospective cohort study of patients ≥18 years of age with confirmed PCN from January 2005 to December 2010 in a community-based integrated care setting in Southern California. Patients with history of acute or chronic pancreatitis were excluded. Malignancy diagnosed within 3 months of cyst diagnosis was considered as pre-existing. Subsequent incidence of malignancy during surveillance was calculated based on person-time at risk. Age- and gender-adjusted standardized incidence ratio (SIR) was calculated with the non-cyst reference population. Recursive partitioning was used to develop a risk prediction model based on cyst imaging features.

RESULTS: We identified 1,815 patients with confirmed PCN. A total of 53 (2.9%) of patients were diagnosed with cyst-related malignancy during the study period. The surveillance cohort consisted of 1,735 patients with median follow-up of 23.4 months. Incidence of malignancy was 0.4% per year during surveillance. The overall age- and gender-adjusted SIR for pancreatic malignancy was 35.0 (95% confidence level 26.6, 46.0). Using recursive partitioning, we stratified patients into low (<1%), intermediate (1-5%), and high (9-14%) risk of harboring malignant PCN based on four cross-sectional imaging features: size, pancreatic duct dilatation, septations with calcification as well as growth. Area under the receiver operator characteristic curve for the prediction model was 0.822 (training) and 0.808 (testing).

CONCLUSIONS: Risk of pancreatic malignancy was lower than previous reports from surgical series but was still significantly higher than the reference population. A risk stratification system based on established imaging criteria may help guide future management decisions for patients with PCN.