11.18 Pancreatic Neuroendocrine Tumor (PNET) Imaging Features are Predictive of Biology

A. Fang1, E. Tashakori1, C. Farinas1, M. Mederos1, A. McElhany1, S. Mohammed1, N. Villifane1, W. E. Fisher1, G. Van Buren1  1Baylor College Of Medicine,Michael E. DeBakey Department Of Surgery, The Elkins Pancreas Center,Houston, TX, USA

Introduction:
Pancreatic neuroendocrine tumors (PNETs) are a heterogeneous group of tumors with variable malignant potential. While most PNETs appear as solid, well-circumscribed, and enhancing masses on computed tomography (CT), their heterogeneous nature can lead to differences in their stereotypical appearance. Therefore, the objective of this study was to test the hypothesis that PNETs with atypical features on CT were associated with more aggressive pathologic features.

Methods:
Through our prospectively maintained Pancreas Surgery Registry, we retrospectively analyzed the radiologic and pathologic features of resected PNETs between January 2005 and December 2015. After independent verification, the CT characteristics such as size, morphology, and enhancement pattern were compared to their histopathologic findings using Chi square and Fisher’s exact tests. Multivariate logistic regression models were generated with backward selection method and a removal p value of 0.1. 

Results:
We analyzed 52 patients who underwent preoperative CT imaging and PNET resection. 29 lesions demonstrated abnormal imaging characteristics such as hypoenhancement (4), isoenhancement (1) calcifications (7), and cystic appearance (19). On univariate analysis, cystic appearance was significantly associated with lower grade and less lymphovascular invasion (p = 0.02, and 0.04, respectively). Bile duct dilation and lymphadenopathy were significantly associated with higher grade and stage, (p = 0.02 and 0.01, respectively). Tumor size significantly correlated with higher stage, positive margin, and lymphovascular invasion (p = 0.02, 0.03, and 0.02, respectively). On multivariate analysis, when controlling for lymphadenopathy, bile duct dilation, and image size, cystic lesions were a significant predictor of lower staging (p = 0.003, 0.01, and 0.01, respectively) and lower rates of lymphovascular invasion (p = 0.04). Hypoenhancment, isoenhancment, and calcifications did not correlate with aggressive pathologic findings.

Conclusion:

This cohort study demonstrated that PNETs with cystic appearance were less aggressive. Conversely, characteristics such as lymphadenopathy, bile duct dilation, and larger lesion size were predictors of aggressive pathologic characteristics. However, enhancement pattern and the presence of calcifications on CT were not associated with more aggressive features.