Building a diagnostic scoring system for high-grade neuroendocrine neoplasms of the pancreas

High-grade neuroendocrine neoplasms of the pancreas (PanNENs), which comprise well-differentiated pancreatic neuroendocrine tumors, grade 3 (PanNET G3s) and pancreatic neuroendocrine carcinomas (PanNECs), are rare but aggressive tumors. Accurate differentiation between PanNET G3s and PanNECs remains a diagnostic challenge, despite their distinct biological behavior and treatment strategies. This study aimed to develop a scoring system to improve diagnostic accuracy using readily available clinicopathologic and immunohistochemical data.

Sixty-six high-grade PanNEN cases underwent clinicopathologic review, immunohistochemistry, and next-generation sequencing. After exclusion of 4 mixed acinar-neuroendocrine carcinomas, 1 diagnostically ambiguous case, and 3 cases with insufficient tissue for next-generation sequencing, 58 cases (29 PanNET G3, 29 PanNEC) were analyzed.

Lasso logistic regression identified predictive features of PanNEC, and multivariable logistic regression was used to assign weights to each factor. Positive predictors of PanNEC included p53 alterations (+4), Rb1 loss (+3), interstitial reaction (+3), co-existing non-neuroendocrine carcinoma (+3), abundant mitoses (+2), and a Ki67 proliferation index greater than 40% (+1). Negative predictors included co-existing PanNET G1/2 (-2), plasmacytoid cells (-1), DAXX/ATRX loss (-1), and somatostatin receptor subtype 2A 3+ (-1). In validation, the average score for PanNEC was 9.52 (median, 10.0); the average score of PanNET G3s was -1.31 (median, -1.0). Using a cutoff of 5.0, the model achieved an area under the curve of 0.989 for distinguishing PanNEC from PanNET G3.

This novel scoring system demonstrated excellent diagnostic performance in differentiating PanNEC from PanNET by integrating morphologic and immunohistochemical features. Prospective studies with larger cohorts are warranted to validate its clinical utility.