ECP 2023 Abstracts

S59 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 a discordant HER2 status between the pre-NAC biopsy and the corre- sponding post-NAC resection, from which 87% (n=165) changed from HER2-0 to HER2-low or vice versa. pCR rate was significantly lower in HER2-low BC compared to HER2-0 BC within the ER+ group (4% versus 5%; p=0.022). In ER- cases, the pCR rate was also lower in HER2-low compared to HER2-0 cases, but not significant (32% versus 34%; p=0.266). There was no statistical difference in overall survival between HER2-low and HER2-0 tumours. Conclusion: Around one third of the BC patients in this study had a discordant HER2 status (0, low, positive) between the pre-NAC biopsy and the post-NAC resection specimen, which could impact clinical decision making if T-DXd gets a role in the treatment of early breast cancer. In this study, HER2-low BC was associated with a lower pCR rate compared to HER2-0 cases in the ER+ group, although these absolute difference was limited and the clinical rel- evance is questionable. Research funded by Aztrazeneca PS-01-010 Alteration of HER2 status during metastatic progression of breast cancer in the HER2-low population K. Bai*, H. Kwon, K. Suh, S. Kim, J. Kim, S. Park *Department of Pathology, Seoul National University Bundang Hospi- tal, Seoul National University College of Medicine, Seongnam, Gyeo- nggi, Republic of Korea Background & objectives: Recent studies show that novel antibody– drug conjugates (ADCs) can improve clinical outcomes of patients with HER2-low breast cancers. This study aimed to investigate the changes of HER2 status during metastatic progression of breast cancer, focusing on HER2-low status. Methods: We analysed 321 paired samples of primary and recurrent/ metastatic breast cancer diagnosed between 2017 and 2022. HER2 status was classified to negative [immunohistochemistry (IHC) score of 0), low (IHC score 1+ or 2+/in situ hybridization-negative) and positive. Discordance rate between primary and matched metastatic samples was calculated in relation to the type of recurrence, site of metastasis and molecular subtype. Results: As a whole, HER2 discordance rate between primary and first recurrent/metastatic breast cancers was 28.0% (k=0.542) with mostly negative-to-low (11.5%) or low-to-negative (10.6%) conver- sion. HER2 discordance between primary tumour and second metas- tasis was 38.5%. Discordance rate was significantly higher in distant metastasis than in loco-regional recurrence (33.2% vs 10.7%). When classified to metastatic sites, discordance rates ranged from 44.4% in bone metastasis to 20.0% in central nervous system metastasis, but with no statistical difference. Luminal A subtype had the highest discordance rate (32.9%) and HER2-positive subtype had the lowest rate (12.1%). Discordant cases revealed significantly higher progester- one receptor (PR) positivity and HER2 negativity in primary tumour, compared to concordant cases. Conclusion: HER2 status changed during metastatic progression in some portion, mostly between negative and low status. HER2 discord- ance between primary tumour and metastases differed by metastatic sites and molecular subtypes and was associated with PR and HER2 status in primary tumour. As HER2 instability increases during meta- static progression, re-evaluation of HER2 needs to be performed when considering ADCs in metastatic settings. PS-01-012 CD47 and CD68 expression in breast cancer is associated with tumour-infiltrating lymphocytes, blood vessel invasion, detection mode and prognosis Y. Chen*, T.A. Klingen, H. Aas, E. Wik, L. Akslen *Fürst medisinske lab., Norway Background & objectives: CD47 expressed on tumour cells binds to SIRPα on macrophages, initiating an inhibition of phagocytosis. We investigated the relation between tumour expression of CD47 and CD68 macrophage content, subsets of tumour infiltrating lymphocytes (TILs), and vascular invasion in breast cancer. Methods: A population-based series of 282 cases from Norwegian Breast Cancer Screening Program was examined. Immunohistochemi- cal staining for CD47 and CD68 was evaluated on TMA-slides. For CD47 evaluation, a staining index (SI) was used. CD68 and TIL subsets (CD45, CD3, CD4, CD8, FOXP3) were counted and dichotomized using immunohistochemistry on TMA-slides. Vascular invasion was determined on whole tissue slides. Results: High CD47 tumour cell expression or high counts of CD68 macrophages were significantly associated with elevated levels of all TIL subsets (p<0.02), CD163 macrophages (p<0.001), blood vessel invasion (p<0.01), and high tumour cell Ki67 (p<0.004). High CD47 expression was associated with ER negativity (p<0.001), HER2 posi- tive status (p=0.03), and interval detected tumours (p=0.03). Com- bined high expression of CD47-CD68 was associated with a shorter recurrence-free survival (RFS) by multivariate analysis (HR 2.37, p=0.018), adjusting for tumour diameter, histologic grade, lymph node status and molecular subtype. Cases with luminal A tumours showed a shorter RFS for CD47-CD68 high cases by multivariate assessment (HR 5.73, p=0.004). Conclusion: This study demonstrated an association of concurrent high CD47 tumour cell expression and high CD68 macrophage counts with various TIL subsets, blood vessel invasion, other aggressive tumour features, and interval presenting breast cancer. Our findings suggest a link between CD47, tumour immune response, and blood vascular inva- sion. Combined high expression of CD47-CD68 was an independent prognostic factor associated with poor prognosis in all cases, as well as in the luminal A category. This work was partly supported by the Research Council of Norway through its Centres of Excellence funding scheme, project number 223250, and by grants from Helse Vest Research Fund (L.A.A.) and Vestfold Hospital Research Fund (Y.C and T.A.K). PS-01-013 The importance of establishing ground truth when building clinical- grade artificial intelligence tools for breast cancer C. Conway*, S. Ali Arshad, P. Khorramshahi, H. Eslami Manoo- chehri, D. Miller, J. Wang, R. Millican-Slater, A. Hanby, J. Stephen- son, D. Bansal, C. Colquhoun, A. Wright, O. Colgan, D. Treanor, C. Salinas *Leica Biosystems, USA Background & objectives: Artificial Intelligence (AI) holds potential to aid pathologists in routine diagnostics and address increasing imbal- ances in pathology demand versus capacity. Herein, we describe the approach and initial results in a study detecting breast cancer using AI. Methods: > 1500 Hematoxylin and Eosin (H&E) slides from mul- tiple global locations were annotated at pixel-level detail. A 25+ breast pathologist network was established ensuring each slide is reviewed/annotated by 3 specialists, accounting for discordance in breast tissue classification into clinically relevant subcategories. Slides are scanned at 40x on Aperio GT 450/450DX scanners and anno- tated using mouse, stylus and touchscreen technology. Results: To develop a ground truth platform for clinical AI devel- opment, consideration should be given to ethical patient slide/data access, sample size, number of specialist reviews for each slide, and variation in tissue preparation/staining. Herein, we describe the results in a study detecting breast cancer using AI. Automatically combining annotation inputs and extracting con- cordant areas is critical to developing a strong data pipeline. Early