Morphologic Variants of Pancreatic Neuroendocrine Tumors: Clinicopathologic Analysis and Prognostic Stratification

Better prognostication/stratification of pancreatic neuroendocrine tumors (PanNETs) is needed. In this detailed morpheomic study of 163 resected PanNETs, 11 unusual variants, some of which were not previously recognized, and others scarcely documented in the literature, were identified, and their pathologic characteristics were further analyzed. By behavior and clinicopathologic associations, these variants could be grouped into three prognostically different categories. I. More aggressive (20%). Included in this group were the variants that in average showed higher grade and stage and adverse outcome including oncocytic, plasmacytoid, lipid-rich and previously unrecognized hepatoid variants, which often had a more diffuse/broad-band growth pattern, with some also displaying discohesiveness. They were characterized by abundant cytoplasm and often had prominent nucleoli (as seen in metabolically active cells), thus the provisional name “metabolic cell phenotype.” Because of their diversion from classical neuroendocrine cytomorphology, these variants created challenges on original diagnostic workup, particularly hepatoid examples, which revealed Arginase 1/Hep Par-1 expression in 50%. II. Less aggressive (10%). These cases either showed signs of maturation, including nested growth, paraganglioid pattern (which was previously unrecognized), and organoid PanNETs such as “ductulo-insular” growth, or showed symplastic/degenerative changes, and despite their paradoxically disconcerting histology, were more benevolent in behavior. III. Undetermined. There were other variants including mammary tubulolobular-like, pseudoglandular, peliotic, and sclerotic PanNETs, which although diagnostically challenging, their biologic significance could not be determined because of rarity or heterogeneous characteristics. Prognostic associations: Features that were significantly different in the more aggressive group than the less aggressive group were median size (5.0 vs 1.6 cm, p < 0.001), percentage of pT3+T4 cases (72% vs 12%, p < 0.001), Ki67 index (5.3% vs 2.3%, p = 0.001), % G2 and G3 cases (77% vs 27%, p < 0.001), and rate of lymph node and distant metastasis (96% vs 27%, p < 0.001). In stepwise logistic regression model using the 3 established prognosticators of T stage, size, and grade along with morphology, only aggressive-morphology (metabolic cell phenotype) was found to be associated with metastatic behavior with an odds ratio of 5.9 with 95% confidence interval (C.I.) 1.688 to 22.945 and p value 0.007. In conclusion, PanNETs display various morphologic patterns that are not only challenging and important diagnostically but appear to have biologic significance. Tumors with more diffuse growth of cells with nucleoli and abundant cytoplasm and/or discohesion (oncocytic, hepatoid, lipid-rich, plasmacytoid PanNETs), provisionally termed “metabolic cell phenotype,” show aggressive characteristics and are an independent determinant of adverse outcome and thus may require closer post-surgical follow-up, whereas variants with more degenerative or mature features (ductuloinsular, pleomorphic, paraganglioma-like) appear to be more benevolent despite their more atypical and worrisome morphology.