Search

Accurate grading of pancreatic neuroendocrine tumors (PanNETs) relies on the assessment of Ki-67 immunohistochemistry (IHC). While digital imaging analysis (DIA) has been employed for Ki-67 IHC assessment in surgical specimens, its applicability to cytologic specimens remains underexplored. This study aimed to evaluate an automated DIA for assessing Ki-67 IHC on PanNET cell blocks. The study included 61 consecutive PanNETs and 5 pancreatic neuroendocrine carcinomas. Ki-67 IHC slides from cell blocks were digitally scanned into whole slide images using Philips IntelliSite Scanners and analyzed in batches using the Visiopharm Ki-67 App in a digital workflow. Ki-67 scores obtained through DIA were compared to pathologists' manual scores. The Pearson correlation coefficient of the percentage of Ki-67-stained nuclei between DIA reads and the originally reported reads was 0.9681. Concordance between DIA Ki-67 grades and pathologists' Ki-67 grades was observed in 92.4% (61/66) of cases with the calculated Cohen's Kappa coefficient of 0.862 (almost perfect agreement). Discordance between DIA and pathologists' consensus reads occurred in 5 PanNET cases which were upgraded from G1 to G2 by DIA due to contaminated Ki-67-stained inflammatory cells. DIA demonstrated excellent concordance with pathologists' assessments, with only minor grading discrepancies. However, the essential role of pathologists in confirming results is emphasized to enhance overall accuracy. • Digital imaging analysis (DIA) of cytology cell block Ki-67 immunohistochemistry (IHC) from pancreatic neuroendocrine tumors • Automated digital workflow to analyze Ki-67 IHC whole slide images • Excellent concordance between DIA and pathologists' consensus scoring of Ki-67 IHC [ABSTRACT FROM AUTHOR]

It is a challenge to accurately assess pancreatic cystic lesions (PCLs) and determine their risk. We compared the yield of tissue acquired with endoscopic ultrasound (EUS)-guided microforceps (through the needle tissue biopsy [TTNB]) with that of samples collected by EUS-guided fine-needle-aspiration (EUS-FNA), and the accuracy of analyses of each sample type in the diagnosis of mucinous PCLs. We performed a prospective open-label study of 114 consecutive adults (56.1% women; mean age, 64.2 y) undergoing EUS-FNA evaluation of PCLs (mean size, 35 mm) at 7 centers, from June 20, 2016, through August 31, 2018. Samples were collected from each cyst by FNA and microforceps; samples collected by FNA were analyzed by cytology and samples collected by TTNB were analyzed by histology. Acquisition yield was defined as the percentage of specimens collected that were adequate for cytologic or histologic analysis. Diagnoses of mucinous cysts were made based on identification of pancreatic mucinous epithelium by cytology analysis of FNA samples or histologic analysis of TTNB samples. Surgical specimens were used as the reference standard when available. The EUS-guided microforceps were successfully inserted into 97.4% (111 of 114) of PCLs. Tissue acquisition yield was significantly higher with TTNB (95 of 114; 83.3%) than FNA (43 of 114; 37.7%) (P <.001). Sixty-one PCLs were determined to be mucinous based on TTNB analysis (53.5%) vs 11 with FNA analysis (9.6%) (P <.001). Among PCLs categorized as equivocal, based on the level of carcinoembryonic antigen, TTNB analysis found 50% (41 of 82) to be mucinous and FNA analysis found 8.5% (7 of 82) to be mucinous (P <.001). Findings from analyses of samples collected by TTNB were 100% concordant with findings from histologic analysis of surgical specimens (14 of 14), whereas only 3 of 14 findings from analysis of samples collected by FNA were in agreement with findings from surgical specimens (21.4%) (P <.001). Four of 5 mucinous PCLs with advanced neoplasia (80%) were detected with TTNB compared with none with FNA (P =.04). Self-limited intracystic bleeding occurred in 7 patients (6.1%), and acute pancreatitis in 6 patients (5.3%). In a multicenter prospective study of patients undergoing EUS-FNA for evaluation of PCLs, we found TTNB collection of tissues for histologic analysis to be safe and feasible, with an acquisition yield of 83.3%. Histologic analysis of samples collected by TTNB identified a larger proportion of mucinous PCLs compared with cytologic analysis of samples collected by FNA—even among samples categorized as equivocal, based on the level of carcinoembryonic antigen. More samples collected by TTNB than FNA were found to have advanced neoplasia. Clinicaltrials.gov no: NCT02979509. [ABSTRACT FROM AUTHOR]