Early detection of CRS after CAR-T therapy using wearable monitoring devices: Preliminary results in relapsed/refractory multiple myeloma (RRMM)

e13626 Background: Chimeric Antigen Receptor T-cell (CART) therapy is almost universally given inpatient due to risks of cytokine release syndrome (CRS). This burdens patients, increases bed occupancy, infection risk and costs. In an investigator initiated clinical trial (IIT), we evaluated the feasibility of using wearable devices for detecting CRS following autologous CART therapy in RRMM, in addition to standard of care (SoC). Reliable early CRS monitoring may help transition to outpatient CART. Methods: The remote wearable device (Current Health Inc.) was worn by patients (pts) as part of an IRB approved IIT from CART infusion to discharge and compared to SoC nursing vital signs. The wearable collected continuous measures of temperature, pulse, respiratory rate and O2 saturation. Clinical CRS timepoints were compared to Current Health (CH) data by 1) 2 clinician independent marking of points where CRS began in CH data plots (31 events) 2) Timestamp when a threshold of 2 standard deviations above each pt’s mean baseline temperatures was breached (12 initial events). Outcomes were time to detection of CRS v/s SoC. Wearable adherence was the duration pts wore the device over the total monitoring period. Results: To date, 16 pts were screened and 14 enrolled (87.5% uptake). 12 of 14 pts experienced CRS with concurrent CH and nursing data. Max CRS grades were 1 (10 pts), 2 (1 pt) and 3 (1 pt). Median length of monitoring was 13 (9-14) days. Wearable adherence during overall monitoring period was 71 (50-81) %, and 88 (69-89) % during high-risk period. By visual inspection method, the wearable detected temperature changes consistent with CRS at a median of 205 (range 50-479) mins earlier than SoC for all events. There was excellent inter-observer correlation (95% CI) = 0.952 (0.901-0.977). Inspection revealed associated variability in pulse, respiratory rate and O2 saturation across all CRS grades, with higher tachycardia and drop in SpO2 noted in Grade 2 and 3 CRS. By threshold temperature method, the wearable detected CRS at a median of 195 (range 16-924) mins earlier than SoC; false positive rate 16%, false negative 0%. Conclusions: Initial results suggest wearables can reliably monitor for CRS in the immediate post-CAR-T infusion period and may facilitate earlier detection. High uptake and adherence reflect pts’ acceptance of the technology. Clearly identifiable deviations in all vital signs on visual inspection, with quantifiable trajectory changes relative to baseline offer a strong foundation for a future machine learning approach to CRS alerting, leveraging network physiology. Augmenting CRS prediction by data from cytokine variations and expanding monitoring to bispecific therapy are planned. Reliable CRS detection by wearables may support the transition from inpatient to outpatient administration of cellular therapies.