The electronic frailty index used to identify high-risk older adults receiving chimeric antigen receptor T-cell therapy.

Abstract

11009 Background: Chimeric antigen receptor T-cell therapy (CAR-T) has transformed outcomes in relapsed/refractory hematologic malignancies and can be effective in older adults. Frailty, a syndrome of vulnerability common in older adults, is associated with poor outcomes but difficult to measure in clinical practice. The relationship between frailty and CAR-T outcomes is poorly understood. Methods: We conducted a retrospective analysis of 260 patients age 65+ years receiving commercial CAR-T at Massachusetts General Brigham (MGB) for multiple myeloma and B cell non-Hodgkin lymphoma. We abstracted patient characteristics and clinical outcomes from the electronic health record (EHR). Using EHR data from the Research Patient Data Registry, we calculated the MGB-eFI, a validated electronic frailty score based upon a deficit accumulation approach to frailty. The MGB-eFI assesses 31 age-related health deficits and yields a score ranging 0-1 (< 0.1 is robust, 0.1-0.2 is pre-frail, 0.2-0.3 is frail, and > 0.3 is very frail). We then evaluated associations of the eFI (as a categorical variable of robust/prefrail versus frail/very frail) with cytokine release syndrome (CRS), immune cell-associated neurotoxicity syndrome (ICANS), overall survival (OS), and event-free survival (EFS) using univariate and multivariable logistic and Cox regression models controlling for covariates. Results: Among 260 patients (median age 73 years [range 65-91], 45% female), the most common CAR-T products used were liso-cel (35%), tisa-cel (18%), ide-cel (18%), cilta-cel (12%), and axi-cel (10%). With the eFI, 10% of patients were robust, 25% prefrail, 28% frail, and 37% very frail. In univariate analyses, frail/very frail was associated with worse OS (hazard ratio [HR] 1.66, p = 0.013), worse EFS (HR 1.78, p = 0.001), higher rates of ICANS (odds ratio [OR] 1.99, p = 0.013), and higher rates of grade 3+ ICANS (OR 4.49, p = 0.003). Frailty was not associated with risk of CRS. In multivariable analyses controlling for age, sex, CAR-T product, lactate dehydrogenase, ferritin, number of prior therapies, and bridging therapy use, frail/very frail by eFI remained associated with worse EFS (HR 1.69, p = 0.005) but not with OS (HR 1.37, p = 0.150). Frail/very frail was numerically associated with increased ICANS, but this did not meet statistical significance (OR = 1.71, p = 0.090). In a multivariable analysis controlling for age, CAR-T product, and ferritin (due to event sample size), frail/very frail was associated with higher rates of grade 3+ ICANS (OR 3.75, p = 0.017). Conclusions: Frailty measured using an automated eFI is associated with worse EFS and higher rates of high-grade ICANS, despite controlling for other clinical and patient factors. The eFI holds promise as a tool that if integrated into the EHR can help clinicians with risk stratification and CAR-T decision-making for older adults with hematologic malignancies.