Abstract PO5-02-02: LyKi1: a highly predictive immune-based score of pathological response to chemotherapy in luminal breast cancer

Abstract

Abstract Background: Identifying which luminal breast cancer (BC) patients will benefit from (neo)adjuvant chemotherapy still remains a challenge. Efficient predictive tests and genomic signatures are still lacking. Because neoadjuvant chemotherapy (NACT) constitutes an in vivo chemotherapy-sensitivity test, we used the pre-treatment core biopsies to identify predictive factors of pathological response in luminal HR+/Her2- BC. Analyses on sTILS and TIL subtypes (CD3, CD4, CD8, CD20) are rare in luminal BC while a predictive value has been reported in TNBC and Her-2 positive BC. Patients and Methods: We performed a retrospective analysis of 211 patients treated with NACT followed by surgery for a T0-T4 luminal (HR+/Her-2-) breast cancer. Association of anthracyclin and taxane was used in 99% of cases. Median age was 48.4 yo (28 – 76). Initial clinical stage was: I: 0.5%, II: 67%; III: 33%. Biopsy analyses were the followings: NST carcinomas: 90%; luminal B: 82%; E&E grade I/II/III: 10%/57%/32%; KI-67 > 20%: 74%. After NACT, RCB 0-1 was observed in 32 cases (15.2%) and pCR in 19 (9.0 %) cases. All pre-therapeutic biopsies were reviewed for pathological factors, blinded to the patients’ outcomes. Clinico-pathological analyses were performed using standard methods. sTILs were estimated on H&E slides. Intratumoral CD3, CD4, CD8 and CD20 infiltrates were carried out by an independent laboratory using a machine learning model for automated and quantitative evaluation based on digital IHC stains. Correlation between sTILs and TIL subtypes were assessed by Spearman’s coefficient. Predictive factors of RCB 0/1 were explored using unconditional logistic regression analysis. After selection on univariate analysis (p< 0.05), multivariate prognostic models were developed using stepwise backward variable elimination process and compared using area under the curve (AUC). Results: sTILs levels ≥ 10% were observed in 15.2% of biopsies. 42 % of tumors had ≥ 10 % CD3-positive cells, 46% for CD4, 38% for CD8, and 8% for CD20. A high level of correlation was observed between these different markers (Spearman’s coefficient ≥ 0.73 for each comparison). By univariate analysis, significant factors (p< 0.05) associated with RCB 0-1 were: age, mitotic index (1 vs 2 vs 3), aneuploidy (1-2 vs 3), differentiation (1-2 vs 3), E&E grade (1-2 vs 3), mitotic count, Ki-67, Magee equation 3, phenotypic group (luminal A vs B), ER, sTILs and each TIL subtype. Age, E&E grade, Ki-67, Magee equation 3, sTILs and all TIL subtypes were selected for multivariate analysis. As TIL subtypes were highly correlated, different predictive models could be performed. The model we selected (called LyKi1) was constructed based on CD8 [OR 2.24 (95%CI 1.55-3.25; p< 10-4)], E&E grade [OR 3.00 (95%CI 1.22-7.42; p=0.017)] and Ki-67 [OR 1.74 (95%CI 1.09-2.76; p=0.019)]. It was highly significantly associated with RCB 0-1, with an AUC value of 0.856 (95%CI 0.756-0.916). For instance, LyKi1 could isolate a group of 30% of patients that achieved RCB 0-1 in 39.7% of cases vs 4.7% for the 70% of patients with lowest LyKi1 scores. pCR rates were 25% and 2%, respectively. Importantly, prediction value of LyKi1 remained very high if restricted to luminal B tumors (AUC 0.817; 95% CI: 0.741-0.894). Ability to predict pCR was also very high (AUC 0.837 (95%CI: 0.750-0.924) in this phenotype). Conclusion: This retrospective analysis clearly highlights the important predictive role of sTILs and TIL subpopulations for the response to an anthracyclin-taxane based regimen in a neoadjuvant setting for luminal breast cancer. Within this cohort, LyKi1 score (combining CD8, grade and Ki-67) has a very high value to predict pathological response to NACT in luminal breast cancer. Importantly, such score has to be validated in a multicentric prospective cohort and investigated in adjuvant setting. Citation Format: Marc Debled, Hugo Deboissy, Coralie Cantarel, Chloé Delfour, Léonie Alran, Marion Fournier, Nathalie Quenel-Tueux, Valérie Velasco, Foucault Chammings, Véronique Brouste, Monica Arnedos, Gaétan MacGrogan. LyKi1: a highly predictive immune-based score of pathological response to chemotherapy in luminal breast cancer [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-02-02.