ECP 2023 Abstracts

S11 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 Conclusion: LTSs display a significantly more immunogenic TME underscoring their effective antitumor immunity, especially at the area of IF compared with STSs. Furthermore, we detected significant intratumoral heterogeneity between TC and IF on the expression of immune determinants, both in LTSs and STSs, which might explain the different antitumor immune responses, affecting patient outcome. The differential expression of immune drivers may help selecting patients for combination therapies to improve antitumor immunity and harness the responsiveness to immune checkpoint inhibitors in PDAC. OFP-03-002 Negative prognostic impact of PD-L1 expression in tumour cells of undifferentiated (anaplastic) carcinoma with osteoclast-like giant cells of the pancreas J. Hrudka*, K. Lawrie, P. Waldauf, V. Ciprová, J. Moravcová, R. Matej *UH Kralovske Vinohrady, Czech Republic Background & objectives: Pancreatic carcinoma is generally charac- terized by a dismal prognosis. A rare subvariant of pancreatic cancer, undifferentiated carcinoma with osteoclast-like giant cells (UCOGC), has an unpredictable prognosis according to many previous studies, with unexpectedly long survival in individual cases. Methods: We examined 13 cases of UCOGCs and performed immu- nohistochemistry focused on the expression of the programmed death-ligand 1 (PD-L1) and several other potential predictive markers (PanTRK, p53, MSH2, PMS2, and the number of tumour-infiltrating lymphocytes), to explore their correlation with the follow-up of the patients. As a control group, we examined 24 cases of conventional pancreatic ductal adenocarcinoma (PDAC). Results: PanTRK was negative in all 24 cases. P53 did not show any significant differences between UCOGCs and PDACs, and the entire cohort was MSH2, MLH1, PMS2, and MSH6 positive. Significant dif- ferences were present in the analysis of PD-L1: UCOGCs were found to express PD-L1 significantly more frequently and have a higher number of tumour-infiltrating lymphocytes than PDAC. The expres- sion of PD-L1 was related to significantly shorter survival in patients with UCOGC and in the entire cohort. Patients with PD-L1 negative UCOGCs displayed surprisingly long survival in comparison to PD-L1 positive UCOGCs and PDACs (both PD-L1+ and PD-L1-). Conclusion: We compared our results with previously published data, and, after statistical analysis, we were able to identify PD-L1 as an effective prognostic marker of UCOGC and suggest a strong need for a clinical trial of immune checkpoint immunotherapy in patients with advanced PD-L1 positive UCOGCs. OFP-03-003 Non-ductal pancreatic tumour classification by whole genome DNA methylation profiling A.V. Verschuur*, F. Westerbeke, J.K. Benhamida, O. Basturk, D.S. Klimstra, S. Yachida, P. Selenica, W.M. Hackeng, C. Geisenberger, L.A. Brosens *Department of Pathology, University Medical Center Utrecht, Utrecht University, The Netherlands Background & objectives: Cytological and histopathological diagnosis of pancreatic neuroendocrine neoplasms, acinar cell carcinoma, solid pseudo- papillary neoplasms and pancreatoblastomas can be challenging while it is crucial for therapeutic decision making. This study investigates if methyla- tion profiling can assist in differentiating non-ductal pancreatic cancers. Methods: DNA methylation profiles were obtained from 306 primary pancreatic cancer samples and 20 normal pancreas cases. Machine learning methods including Neural Networks, Random Forest and Gradient Boosting were trained to distinguish between cancer types. Methylation data obtained from The Cancer Genome Atlas spanning 29 different tumour types were used to develop an algorithm capable of detecting tumours of non-pancreatic origin. Results: The Neural Networks, Random Forest and Gradient Boosting models achieved accuracies of 96%, 9% and 91% respectively in 100% of cases. When modifying the threshold for minimally required prob- ability scores for classification, the Random Forest model demonstrated highest accuracies for the majority of cases. Non-pancreatic tumour detection based on prediction scores achieved an AUC of > 0.99. Conclusion: Highly accurate differentiation between non-ductal pan- creatic cancers can be achieved with whole genome DNA methylation profiling. At the same time, non-pancreatic entities are identified with near perfect precision preventing false positive diagnoses. OFP-03-004 Morphological spectrum of gallbladder cancer and its precursors V. Angerilli*, T. de Bitter, M.E. Vink-Börger, P.R. de Reuver, M. Fassan, I.D. Nagtegaal, R.S. van der Post *Department of Medicine, Surgical Pathology Unit, University of Padua, Italy Background & objectives: Gallbladder cancer (GBC) is a rare neo- plasm in Western countries with a dismal prognosis and its pathological evaluation remains a challenge. Here, we performed a comparative histopathologic analysis of a large multi-institutional series of GBC cases and associated precursor lesions. Methods: Specimens and clinicopathological data of 670 GBC patients, diagnosed between 2000 and 2019 were collected using the Dutch Nationwide Pathology Database (PALGA) and the Netherlands Cancer Registry. Histopathology of precursor lesions and carcinomas was reviewed. The entire cohort was investigated for the immunohisto- chemical expression of EMA, MUC2, MUC5AC, MUC6, CK7, CK20, and p53. Univariate and multivariate survival analyses were performed. Results: Precursor lesions were reported in 41.8% (280/670) of cases, either intracholecystic papillary-tubular neoplasm (ICPTN) (n=192), biliary intraepithelial neoplasia (BilIN) (n=81), or mixed lesions (n=7). Histopathological subtypes comprised biliary-type adenocarcinoma (AC, 64.8%), intestinal-type AC (13.6%), and other subtypes (21.6%) including adenosquamous, diffuse, undifferentiated, mucinous, neu- roendocrine, squamous, clear cell and carcinosarcoma. The presence of BilIN was associated with biliary-type AC (p=0.040), while ICPTN was associated with intestinal-type AC (p=0.002). At univariate analy- sis, carcinoma histotype, grading, T and N stage, lymphatic, vascular, and perineural invasion were associated with overall survival (OS) (p<0.05). At multivariate analysis, carcinoma histotype, grading, and T stage proved to be independent prognostic factors (p<0.05). Conclusion: This work sheds light on the morphologic spectrum of GBC and precursor lesions. The histopathologic and immunopheno- typic features of GBC represent valuable information and must be taken into account in the prognostic stratification of GBC patients. OFP-03-005 Diagnostic accuracy of cell blocks in bile duct carcinomas. A pilot study L.P. Mocan*, I. Rusu, T. Mocan, C.S. Melincovici, C.M. Mihu *Department of Histology, “Iuliu Ha ț ieganu” University of Medicine and Pharmacy, Cluj-Napoca, Romania Background & objectives: Cholangiocarcinoma (CCA), a highly aggressive malignancy has a rising incidence and a deficient diagnos- tic accuracy, of around 50%. The aim of this study was to assess the diagnostic performance of cell blocks from bile samples in patients with biliary carcinomas.