AJSP2603 - CMLE - AJCP CME: Building a diagnostic scoring system for high-grade neuroendocrine neoplasms of the pancreas-AJCP CME: Building a diagnostic scoring system for high-grade neuroendocrine neoplasms of the pancreas

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

Pdf Summary

This multi-institutional study developed and validated a practical diagnostic scoring system to distinguish the two main types of high-grade pancreatic neuroendocrine neoplasms (PanNENs): well-differentiated pancreatic neuroendocrine tumor grade 3 (PanNET G3) and poorly differentiated pancreatic neuroendocrine carcinoma (PanNEC). Accurate separation is clinically critical because the entities differ in biology and first-line treatment (targeted/somatostatin-based therapies and surgery for PanNET G3 vs platinum-based chemotherapy for PanNEC), yet they share the same WHO proliferative thresholds, making morphology alone challenging—especially in small biopsy or EUS-FNA samples.<br /><br />From 66 initially diagnosed high-grade PanNENs, the authors excluded mixed acinar–neuroendocrine carcinomas, one persistently ambiguous case, and cases lacking sequencing tissue, leaving 58 tumors (29 PanNET G3, 29 PanNEC). Cases underwent blinded morphologic review, focused immunohistochemistry (including p53, Rb1, Ki67, DAXX/ATRX, and SSTR2A), and limited next-generation sequencing in a subset.<br /><br />Using Lasso logistic regression for feature selection and multivariable modeling for weighting, the final point-based score incorporated readily assessable predictors. Features favoring PanNEC were p53 abnormality (+4), Rb1 loss (+3), interstitial reaction (+3), coexisting non-neuroendocrine carcinoma (+3), high mitotic activity (+2), and Ki67 >40% (+1). Features favoring PanNET G3 were coexistence with PanNET G1/2 (–2), plasmacytoid cells (–1), DAXX/ATRX loss (–1), and strong SSTR2A expression (score 3) (–1).<br /><br />With a cutoff score of 5, the system showed excellent discrimination (AUC 0.989; accuracy 0.966). The authors conclude the algorithm improves reproducibility and is particularly useful when tissue is limited, though larger prospective validation is needed.

Keywords

high-grade pancreatic neuroendocrine neoplasms

PanNET G3

PanNEC

diagnostic scoring system

Lasso logistic regression

immunohistochemistry markers

p53 abnormality

Rb1 loss

Ki67 proliferation index

SSTR2A expression