ECP 2023 Abstracts

S36 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 yr, 3.3 cm) than FA (37.7 yr, 2.4 cm) (p=0.02). Logistic regression model showed older age and larger tumour size were strong predic- tors of PT. BIRAD, needle gauge, clinical presentation, and history of breast cancer were not associated with upgrade to PT. Conclusion: Core biopsy is sufficiently sensitive and specific in strati- fying FELs of the breast and can effectively optimize and reduce the need for surgical management in FELs. In indeterminant core biopsies, pertinent clinical information such as age and tumour size can poten- tially aid in appropriate decision making. Core biopsy plays a crucial role in safe avoidance of unnecessary surgical procedures of FELs of the breast, given the vast majority of FAs can be observed. OFP-09-010 The assessment of tumour-infiltrating lymphocytes in invasive apo- crine carcinoma of the breast in relation to the Her2 status S. Vranic*, N. Kuzumova, I. Rose, Z. Gatalica *College of Medicine, Qatar Background & objectives: Tumour-infiltrating lymphocytes (TIL) represent a robust immunologic biomarker mediating chemotherapy and immunotherapy responses. In the current study, we assessed the tumour-infiltrating lymphocytes (TIL) in invasive apocrine carcinomas of the breast (Apoca) in relation to the Her2 status. Methods: The current study used a cohort of 54 well-characterized, naïve Apoca. All patients were women, with a mean age of 64 years (range 39–83 years). The TIL distribution was assessed using the hematoxylin and eosin whole slide/scanned images following the recommendations of the interna- tional TILs working group 2014 [Ann Oncol 26(2), 2015]. Results: Forty carcinomas were “pure apocrine” [PAC; ER-/AR+)], and the remaining 13 were classified as “apocrine-like” [ALC; ER+/-, AR+/-]. HER-2/neu was positive (score 3+ by IHC and/or amplified by FISH) in 18/40 (45%) PAC and 3/13 ALC (23%). The prevalence of HER2-low expression (scores 1+ and 2+ without HER2 amplifica- tion) in Apoca was high (21/53, 40%); the HER2-low phenotype was more prevalent in “triple-negative” PAC than in ALC (68% vs. 46% p<0.001). Levels of TIL were low (≤10%) in 75% Apoca (median: 5%, range 0–50%). TIL levels were significantly higher in ALC than in PAC (p = 0.05). Her2 status had no impact on the TIL distribution (p=0.45). Conclusion: Invasive apocrine carcinomas of the breast have predomi- nantly low TIL, particularly PAC. The prevalence of the HER2-low phenotype in apocrine carcinomas is high, which should have thera- peutic and clinical implications given the recently approved therapies with antibody-drug conjugates for Her2-low breast cancers. OFP-09-011 Heterogenous HER2 inter- and intra-laboratory lower limit of detection N. ’t Hart*, S. Sompuram, M. Eenkhoorn, D. Dabbs, S. Bogen *Isala, The Netherlands Background & objectives: HER2 (4B5) immunohistochemistry (IHC) was used in the Destiny/Breast04 study. Accurate patient stratifica- tion depends on whether the assay can be consistently replicated from one lab to the next. The objective is to quantitatively measure inter- laboratory reproducibility using HER2 calibrators. Methods: To quantify reproducibility of HER2 IHC assays, three calibrator slides (Boston Cell Standards Inc, Boston, USA) were sent to 42 labs. The calibrator incorporates 10 HER2-coated cell-sized microbeads at HER2 concentrations ranging 38.000 – 2.7 million copies. Linear range formalin- fixed microbead controls were also included. Slides were stained on three consecutive days, analysed and the coefficient of variation was calculated. Results: Four antibodies were used by 32 labs: 59.4% clone 4B5, 9.4%Her- ceptest monoclonal (DG44), 15.6% SP3 and 15.6% C-erbB-2. The calibra- tor was evaluable for 4B5 and DG44 (22 laboratories), allowing calculation of the lower limit of detection (LOD). The inter-laboratory coefficient of variation (CV) for the LODs was 55% (4B5 labs) and 5% (DG44 labs). In addition to LOD, we also evaluated the CV in stain intensity with a mid- range control: 32% (4B5) and 22% (DG44). Within-lab variability in LODs were: 25% (4B5) and 18% (DG44). Within-lab variability with a single mid-range control was 23% (4B5) and 12% (DG44). Conclusion: HER2 IHC assays are more heterogenous than commonly appreciated, sometimes with poor inter- and intra-laboratory reproduc- ibility. DG44 showed the highest consistency, followed by 4B5. The fact that there were only 3 participating IHC labs using the DG44 is a limitation in the findings. Laboratories need to re-validate HER2 IHC assays to ensure accurate identification of HER2-low and address poor reproducibility issues in order to reliably assess HER2-low in clinical practice. OFP-09-012 A multi-feature AI solution for diagnosis support of breast exci- sions: a clinical validation study G. Broeckx*, F. Deman, R. Salgado, I. Krasnitsky, D. Mevorach, J. Theunissen, M. Vecsler *GZA-ZNA Antwerp, Belgium Background & objectives: This study aimed to clinically validate the performance of an artificial intelligence (AI)-based solution on detec- tion of invasive and in-situ carcinomas in breast excisions compared to a robust ground truth (GT) established by multiple expert pathologists. Methods: An AI algorithm previously validated on biopsies was tested in a prospective standalone performance study on an external cohort comprising 248 retrospectively collected breast excisions. AI results on invasive and in-situ carcinoma were compared with GT diagnosis that was reached by concordance between two blinded pathologists who prospectively reviewed the slides. Discrepancies were adjudicated by a third expert pathologist. Results: The AI demonstrated high performance when compared with the GT with an AUC of 0.986 (95% CI: 0.973-0.998) for the detec- tion of invasive carcinoma (specificity and sensitivity of 96.3% and 89.9% respectively) and with an AUC of 0.994 (95% CI: 0.987-1) for the detection of DCIS (sensitivity and specificity of 95.6% and 95%, respectively). The AI differentiated well between subtypes/grades of invasive and in-situ cancers with an AUC of 0.963 (95% CI: 0.922-1) for IDC vs. ILC and AUC of 0.970 (95% CI: 0.931-1) for DCIS high grade vs. low grade. Conclusion: This blinded study reports the successful clinical validation of an AI-based solution in the accurate detection of clinically relevant diagnostic parameters regarding invasive and in situ breast carcinoma, offering an important tool for computer-aided diagnosis in routine pathol- ogy practice, supporting pathologists in their diagnostic work. This research is sponsered by Ibex Medical Analytics. OFP-09-013 Interplatform and interobserver reproducibility of PD-L1 CPS testing in mTNBC using 22C3 and SP263 assays M. Ivanova*, C. Frascarelli, B. Cerbelli, G. Pignataro, C. Crisci- tiello, G. Curigliano, K. Venetis, E. Sajjadi, E. Guerini Rocco, P. Graziano, M. Martini, G. D’amati, N. Fusco *IEO, European Institute of Oncology, Milan, Italy Background & objectives: Pembrolizumab+chemotherapy is a first-line treatment option for metastatic triple-negative breast cancer (mTNBC) with PD-L1 combined positive score (CPS)≥10. The avail- able validated assays lack standardization. We aimed to characterize the reproducibility of PD-L1 CPS in mTNBC using CE-IVD tests. Methods: 60 mTNBC samples were tested for PD-L1 IHC using 22C3 pharmDx on a Dako Autostainer Link 48 and SP263 on a Ventana BenchMark Ultra. Five pathologists assessed the CPS. The intraclass correlation coefficient (ICC) for each assay and inter-rater/inter-platform