ECP 2023 Abstracts

S121 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 Conclusion: Wnt proteins have multiple functions in cell proliferation, migration, and tissue organization. This study showed that the main source of Wnt3α in the lungs of naïve mice are club cells within the bronchial epithelium. Post-bleomycin challenge, club cells within bron- chial epithelium are recovering but no longer express Wnt3α in the same amount. Since a small fraction of M2 macrophages expresses Wnt3α in a later stage post-bleomycin challenge, their exact role has to be elucidated further. PS-20-009 Comparison of PDL1 expression between different tissue samples in non-small cell lung cancer Z. Radicevic*, N. Kurtović, I. Zagorac, M. Bakula, E. Matuzalem Marinović, J. Rajc *Department of Pathology and Forensic medicine, Osijek University Hospital, Osijek, Croatia; Faculty of Medicine Osijek, University of Osijek, Croatia Background & objectives: In many patients with advanced stage of non-small cell lung cancer, only small biopsies or cytology samples are available for the evaluation of PDL1 expression. In this study we compared the average rate of PDL1 positivity between different tissue samples. Methods: We reviewed 401 case of non-small cell lung cancer stained with PDL1 IHC Ventana SP263 assay between January 2019 and December 2022, assessed as a tumour proportion score (TPS). Sam- ples were divided in four categories: cell blocks, small biopsy samples, surgical resections and distant metastasis. The cross-tabulated statistics of the observed values were performed between the four categories. Results: Among the 401 specimens, 187 were small biopsies (46,6%), 130 cell blocks (32.4%), 51 lung resections (12,7%) and 33 biopsies taken from distant metastasis (8,2%). Tumour proportion score (TPS) of <1% was detected in 205 cases (51,1%), TPS ≥1-49% in 133 cases (33,2%) and TPS ≥ 50% in 63 cases (15,7%). By using both, continu- ous and three-category variables of TPS (<1%, ≥1-49% and ≥50%) we found that the average PDL 1 positivity rates were evenly distributed in all examined tissue samples (cell blocks, small biopsies, large surgical resections, and distant metastasis), considering all three TPS categories. Conclusion: Our study shows that the rates of PDL1 positivity were not significantly different in cell blocks, small biopsy samples, samples from distant metastasis and large surgical resections. Despite of the well know heterogeneity of PDL 1 expression in non-small cell lung cancer that might reflect in inaccurate results, especially if the test is carried in a small tissue specimen, our findings suggests that all above- mentioned materials are equally adequate for immunohistochemical PDL 1 assessment in a daily practice. PS-20-010 Real-world comparison between PD-L1 SP263 and 22C3 assays in non-small cell lung cancer: interchangeability of two assays and interlaboratory concordance M.S. Roh*, S.Y. Ha *Department of Pathology, Dong-A University College of Medicine, Republic of Korea Background & objectives: The evaluation of PD-L1 expression is critical for the selection of patients for immunotherapy in non-small cell lung cancer (NSCLC). To standardize the effective use of PD-L1 assays, inter-assay concordance for two PD-L1 tests was assessed using real-world data. Methods: Two histologic sections from 340 NSCLCs were distributed to 2 sites. SP263 testing on Ventana Benchmark platform was performed in- house and interpretated by a pathologist, while 22C3 testing on Dako Link 48 platform was outsourced and evaluated by 10 different pathologists. The concordance rate was evaluated using tumour proportion score (TPS) cutoffs of 1 and 50, respectively. Results: The overall percentage agreement (OPA) was 89.6% at TPS ≥ 1, whereas the OPAwas 98.1% at TPS ≥ 50. The κ coefficient was calculated as 0.653 at TPS ≥ 1 and was 0.864 at TPS ≥ 50. The majority of scor- ing challenging cases was in TPS 1% borderline samples. When examin- ing variability between pathologists, there was good to strong reliability among pathologists at TPS ≥ 50 (intraclass correlation coefficient [ICC] = 0.78–0.94), whereas moderate-to-strong interobserver agreement (ICC = 0.51-0.81) at TPS ≥ 1. The interobserver reproducibility was improved in the recent 2-year samples compared to the previous samples. Conclusion: These findings indicated that PD-L1 SP263 and 22C3 assays were highly comparable analytical performance, especially at TPS ≥ 50, although assays were performed at different laboratories and evaluated by differently trained pathologists. Experience gained with real-world analysis supported the potential interchangeability of two PD-L1 assays. By reduc- ing the number of PD-L1 assays in cases displaying an unequivocally TPS ≥ 50, it may aid to save both cost and specimens for subsequent molecular evaluations. PS-20-011 Comparison of artificial intelligence-assisted and manual assessment of programmed death-ligand 1 (PD-L1) expression Z. Sağnak Yılmaz*, Z. Turkmen Usta, S. Aydin Mungan *Dokuz Eylül University, The Institute of Health Sciences, Department of Molecular Pathology, Karadeniz Technical University, Faculty of Medicine, Department of Pathology, Turkey Background & objectives: Programmed cell death ligand 1 (PD-L1) is a critical biomarker for predicting the response to immunotherapy. In this study, we aim to evaluate the agreement in PD-L1 scoring among pathologists and between deep learning (DL)-based artificial intelli- gence (AI) model. Methods: One hundred five patients with lung cancer were included in the study. Preparations stained with Ventana PD-L1 (SP263) assay were first evaluated by three pathologists and given tumour propor- tion score (TPS). All preparations were evaluated with Ventana DP 200 slide scanner and uPath/ PD-L1 algorithm. Statistically, the intra- class correlation test and kappa score were used to evaluate group consistency. Results: Considering the agreement among pathologists (P1, P2, P3), the highest correlation was between P1 and P2 (P1 vs. P2: R= 0.894; P2 vs. P3: R= 0.879; P1 vs. P3: R= 0.794). Considering the correla- tions of TPS-AI with pathologists separately, the highest accordance was with P3 (TPS-AI vs. P3: R= 0.772; TPS-AI vs. P2: 0.769; TPS-AI vs. P1: 0.657). Those with TPS <1%, 1-49%, and >50% were cat- egorized separately. The agreement between TPS-AI and pathologists is generally lower than inter-pathologists’ agreement (Kappa values: TPS-AI vs. P2: 0.537, TPS-AI vs. P1: 0.500, TPS-AI vs. P3: 0.482; P1 vs. P2: 0.739; P2 vs. P3: 0.702; P1 vs. P3: 0.554). Conclusion: In general, both the agreement among pathologists and the agreement of individual pathologists with TPS-AI are lower than the literature data. This may be due to the limited number of cases evaluated. It was concluded that manually given scores were higher than TPS-AI. The most important reason for this situation may be the pressure to approach the assumed threshold value for the patient to receive immunotherapy. PS-20-012 Confocal microscopy and real-time PCR for rapid perioperative molecular profiling of lung cancer: a proof of concept F. Schmitt*, A.G.M. Sciarra, M. Christodoulou, I. Letovanec, C. Beniere *SamanTree Medical, Lausanne, Switzerland