Despite therapeutic improvements, resistance to palbociclib is a growing clinical challenge which is poorly understood. This study was conducted in order to understand the molecular mechanisms of resistance to palbociclib, and to identify biomarkers to predict who will take advantage from cyclin-dependent kinase 4/6 inhibitors (CDK4/6i). A total of about a thousand blood samples were collected from 106 patients with hormone receptor positive (HR+) human epidermal growth factor receptor 2 (HER2) negative metastatic breast cancer who received palbociclib in combination with fulvestrant as the first-line metastatic therapy enrolled in this study. The genotyping of their plasma cell-free DNA was studied, including serial plasma samples. Collectively, our findings identify the appearance of KRAS mutations leading to palbociclib resistance acquisition within 6 months, and provide critical information for the prediction of therapeutic responses in metastatic breast cancer. By monitoring KRAS status through liquid biopsy, we could predict who will take advantage from the combination of palbociclib and fulvestrant, offering highly-individualized treatment plans, thus ensuring the best patient quality of life.

Introduction Breast cancer is a heterogeneous disease. Our study focuses on a monoinstitutional series of patients affected by Hormone Responsive carcinomas (luminal A and luminal B) and aims to define an optimal Ki-67 cut-off, to correctly stratify these patients into risk classes, using the ImmunoHistoChemical (IHC) surrogates of the Molecular Subtypes, according to the St. Gallen guidelines. Methods We analyzed 1685 patients. These patients underwent both radical and conservative surgeries with Sentinel Lymph Node Biopsy eventually followed by Axillary Dissection (AD). Furthermore, all the patients underwent adjuvant therapies according to the guidelines. A retrospective univariate analysis was performed and survival curves (Disease-Related Survival, DRS, and Disease-Free Survival, DFS) were carried out according to the following ki-67 risk classes: Low Risk (Ki-67 ≤ 14%); Intermediate Risk (Ki-67 15% ÷ 20%); High Risk (Ki-67 > 20%). Results 14 yy DRS was 98% in LA and 85% in LB with a ki-67 cut-off of 14% (p=0.037) vs 95% (LA) and 83% (LB) with a ki-67 cut-off of 20% (p=0.003). 14yy DFS was 85% in LA and 72% in LB with a ki-67 cut-off of 14% (p=0.017) vs 83% (LA) and 66% (LB) with a ki-67 cut-off of 20% (p<0.000). Discussion Our results confirmed that the 20% Ki-67 cut-off is more reliable in differentiating patients at low or high risk of recurrence and death, and stratifying patients eligible for adjuvant chemotherapy. Thus, despite its poor reproducibility, the identification of the most accurate ki-67 index assumes a pivotal relevance in guiding a tailored strategy among patients with this specific profile of breast cancer, as well as the molecular surrogates, in order to avoid harmful overtreatments.

Background: The number of early clinical trials evaluating immunotherapies (IT) has increased exponentially over the last decade. These agents expose to adverse events (AE), which have been largely described with immune checkpoint blockers (ICB) as a monotherapy. However, the safety profile of IT other than ICB or in combination, have been poorly characterized. Here, we comprehensively analyze the safety profile of IT in phase I/II (P1/2) trials when used as a monotherapy (M) or in combination (C) with other anti-cancer agents, and compare it to the one of other anticancer agents assessed in the same patient (pt) population. Methods: All consecutive pts treated in a P1/2 trial at the Gustave Roussy Drug Development Department between Jan 2008 and Aug 2020 were included. Pt and trial characteristics, AEs of any grade (G) and any type occurring at any cycle on trial, dose administered and treatment modifications were described in the following subgroups: IT as a M (IT-M), targeted therapy (TT) as a M (TT-M), IT in C (IT-C), TT in C (TT-C), other.

507 Background: Despite improvements in treatments, patients diagnosed with TNBC still have poor prognosis for a higher tendency of developing BrM. Identifying patients at high risk of BrM, enabling to predict who will take advantage from appropriate additional treatment, remains a critical problem. ctDNA represents a valuable tool associated with the outcome and the aggressiveness of breast cancer but no prognostic and predictive biomarker has been identified to predict the development of BrM in TNBC. We studied the usefulness of assessment of CSF-ctDNA for early identification of the risk of BrM in TNBC. Methods: Between January 2016 and December 2020, 323 newly diagnosed non-metastatic TNBC patients who underwent neoadjuvant therapy+surgery(NACT) with complete response(CR)were prospectively enrolled. After surgery, samples of CSF measuring ctDNA were obtained from all patients: CSF-ctDNA was extracted with the QIAamp Circulating Nucleic Acid Kit (Qiagen, Valencia, CA, USA) and ctDNA levels were measured. Survival curves were estimated using the Kaplan-Meier method and compared with the Log-rank test. Multivariate Cox regression was used to identify the risk of mortality at three years. Results: After NACT, CSF-ctDNA was detectable in 126/323 (39%) patients, 101/126 (80%) were diagnosed at III stage. 124 of 126 (98.4%) ctDNA+ patients subsequently developed BrM. In contrast, only 2 (2/197, 1%) ctDNA- patients subsequently developed BrM and the 195 other patients remain in a CR (p < 0.001, Fisher's exact test). CSF-ctDNA did associate with PFS and OS: undetectable ctDNA was associated with superior PFS (HR 0.3; p = 0.002) and OS (HR 0.2; p < 0.01), indicating survival is largely determined by the onset of BrM. With a median follow-up of 3 years, median PFS of ctDNA+ vs ctDNA- patients was 13 months vs not reach, p = 0.004 (by Log-rank test). Median OS for ctDNA+ vs ctDNA- patients was 16 months after NACT vs not reach, p = 0.0016 (by Log-rank test). At multivariate analysis detectable CSF-ctDNA emerged as the best predictor of the develop of BrM and 24-month mortality (HR:3.62; p < 0.0001). Age, stage, Ki67% and response to chemotherapy were not significantly associated with the prognosis. Conclusions: After NACT, detectable CSF-ctDNA significantly associates with PFS and OS, identifying early at-risk patients to develop BrM in TNBC.

1053 Background: Palbociclib (P) in combination with fulvestrant (F) or letrozole (L), is used globally to treat metastatic breast cancer but despite therapeutic improvements most patients acquire resistance to CDK4/6i. KRAS tumor mutations (mutKRAS) have been associated with worse PFS in several tumor types but have not been analysed extensively in breast cancer. To understand the molecular mechanisms of resistance to CDK4/6i and their clinical behavior, using liquid biopsy, we evaluated the opportunity to reveal the onset of resistance to CDK4/6i detecting mutKRAS ctDNA. Methods: We studied the KRAS mutation status of 211 patients with mBC treated with CDK4/6i plus L or F as first-line metastatic therapy. Using Bio-Rad QX200 droplet digital polymerase chain reaction (ddPCR) system we determined KRAS ctDNA levels in plasma. Using logistic and Cox regression, a predictive model for objective response (OR), progression-free survival (PFS) and overall survival (OS) was constructed. The PFS and the OS were estimated by the Kaplan–Meier method and compared with use of the log-rank test. Results: In 38% (81 patients, 24 in treatment with L and 57 in treatment with F) we observed mutKRAS ctDNA before starting CDK4/6i: the detection of mutKRAS significantly correlated with the onset of resistance to CDK4/6i within 6months from the evidence of KRAS mutation and worse PFS ( p<0.001). OR was seen in 84 of 130 KRAS wild-type (WT) patients versus 0 of 81 in KRAS mutants. At 24-month follow up, median PFS was significantly better in KRAS WT versus mutants (3.1 [range: 1-6months,95%CI 0.9-3.6] versus NA months; p<0.001). Correlating the results of liquid biopsy both to tumoral burden and patients clinical features, we observed a higher mutKRAS circulating copies-number in those patients with two or more metastatic sites( p<0.001). Conclusions: Despite the study’s limitations, our data suggest mutKRAS ctDNA status leads to CDK4/6i resistance acquisition within 6 months from the detection and provide critical information for the prediction of therapeutic responses in mBC. Monitoring KRAS status with liquid biopsy, we could predict who will take advantage from CDK4/6i, decreasing wastes of resources ensuring the best patients’ quality of life.