Abstract P4-02-01: Clinical and biological markers to improve the identification of frailty in patients with early breast cancer

Background: Aging is associated with deficits accumulation and increased vulnerability to adverse outcomes, ultimately resulting in frailty. Comprehensive geriatric assessment (CGA) is considered the gold standard to assess aging. Using the information gathered as part of the CGA in elderly breast cancer (BC) patients (pts), our group has designed a 43 item-frailty index (43-FI). Over the last few years, the analysis of cellular senescence through senescence-associated secretory phenotype (SASP) has been suggested as alternative biomarker of aging. In this study, we identified a panel of 23 SASP-related factors(RF) that was correlated with the 43-FI, and both with prognosis, resulting in new indicators for aging assessment. Methods: 164 pts aged ≥65 years, with early BC included in 3 onco-geriatric trials, were evaluated by the means of a CGA. Geriatric assessments and blood sampling were performed after breast surgery, and before the initiation of systemic treatment (adjuvant endocrine therapy or chemotherapy). Serum samples were available for 140 pts, and were analyzed to identify the 23 SASP-RF. Pts were classified as fit, vulnerable or frail according to the Balducci criteria (Balducci-c) and the 43-FI. The 23 SASP-RF were correlated with the 43-FI and with survival. For each SASP-RF, Area Under the Receiver Operating Characteristic Curve (AUC-ROC) analysis on the fit and frail pts was perfomed to identify the best performing markers (AUC > 0.65) and optimal cut-points. Marker-specific optimal cut points were then used to classify vulnerable pts into high- and low-risk groups. A senescence score (SenS) was developed by training a Random Forest model on fit and frail pts considering the best-performing SASP-RF, defined as those showing relative importance > 5%. The SenS was used to classify vulnerable pts into high- and low-risk groups. Overall survival (OS) was computed from study entry to death from any cause. Results: The 43-FI identified 20% of pts as fit, 71% as vulnerable, and 9% as frail. Vulnerable and frail pts were older than fit (77 and 78 yrs respectively vs 73 yrs), and had less aggressive tumors (62% and 60% of vulnerable and frail pts had Ki67 < 20%; 71% and 80% had G1-2). Hormone receptor-positive HER2-negative tumors were predominant in all three categories. Chemotherapy was administered more often to fit (63%) than vulnerable (33%) and frail (7%) pts. 43-FI and Balducci-c provide markedly concordant classification of pts (frail pts identified by Balducci-c had higher 43-FI score, overall percent agreement= 47%). The 43-FI identified pts with different prognosis: median(m)-OS was 5.5 months (mo) in frail vs 13 mo in fit (HR=4.39, 1.88-10.23 95% CI, p=0.001), and 9.8 mo in vulnerable pts. This association was independent from age. A total of 8 SASP-RF were found associated with the 43-FI, including GDF15, ICAM1, TIMP1, HUIL9, HUGCSF, HUMIP1A, HUMIP1B, HU TNFA. Among these, GDF15 was the best associated with frailty (AUC=0.88, fit vs. vulnerable p=0.001; Fit vs. Frail p=0.001; Vulnerable vs. Frail p=0.038). Applying GDF15 optimal cut-point to the vulnerable cohort, we find that higher levels of GDF15 were correlated with worse prognosis (mOS 11.6 mo low-GDF15 vs 7.8 mo high-GDF15, HR 2.23, 1.29-3.85 95% CI, p=0.003). The SenS model, including 7 SASP-related markers (GDF15, TIMP1, ICAM 1, HU IL9, HU MIP1A, HU MIP 1B, TNF a), was positively associated with prognosis (mOS 10.9 mo low-SenS vs 9.3 mo high-SenS, HR 1.97, 1.14-3.41 95% CI, p=0.013). Conclusions: Our findings suggest that the 43-FI may serve as a potential clinical marker of aging in pts with early BC. GDF15 and SenS, could be used to dissect the heterogeneity of vulnerable pts, identifying subgroups with different prognoses within this cohort. Further analyses in a wider population are needed to confirm these results.