High-Frequency Temperature Monitoring Detects Clinically Actionable Fevers Otherwise Missed By Standard-of-Care Monitoring during Treatment of Hematologic Malignancies

Introduction. Fever is an important marker of treatment-related adverse events in hospitalized patients undergoing treatment for hematologic malignancies. The use of non-invasive, wearable sensors for high-frequency temperature monitoring (HFTM) has been demonstrated to detect adverse febrile events earlier and more frequently than conventional, standard-of-care (SOC) nursing vitals. Given the increased sensitivity of event detection that is inherent with HFTM, a formal methodology for characterizing the clinical etiology of device-detected fevers becomes increasingly important for the development of clinical decision-making algorithms for these emerging technologies.

Methods. We conducted a prospective observational study measuring high-frequency temperature readings every 2 minutes using a self-administered wearable device (TempTraq, BlueSpark Technologies, Inc.) in 62 hospitalized patients undergoing either hematopoietic stem cell transplant (HCT) or chimeric antigen receptor T-cell (CAR-T) therapies. Corresponding HFTM and nursing SOC data streams were then directly compared to categorize fever events as either concordant (detected by both HFTM and SOC) or discordant (detected by either HFTM or SOC alone). A retrospective review of the medical record was performed to identify significant clinical events associated with each febrile event, including therapy-related toxicity, sources of infection, medication side-effects, blood product transfusion reactions, and thrombotic events.

Results. PreliminaryHFTM data collected from the HCT cohort (n= 39) and CAR-T cohort (n=23) resulted in the capture of 421,367 unique temperature readings compared to 4,816 readings by SOC nursing vitals over a cumulative monitoring period of 585 days. Of 58 and 42 febrile events detected by HFTM and SOC monitoring respectively, 39 fevers were found to be concordant, 3 febrile events were detected by SOC vitals alone, and 19 fevers were detected by HFTM but missed by SOC vitals. Blinded retrospective review resulted in the assignment of a putative fever etiology to 15 of 19 HFTM-only febrile events, with 7 of these events being potentially clinically actionable (antibiotics for infection, management of treatment toxicity, etc.) had they been identified in real time.

Conclusion. Most febrile events detected by HFTM alone in the preliminary results of this study could be attributed to an underlying putative clinical etiology, some of which would have been clinically actionable had they been identified in real time. This suggests that incorporation of HFTM into routine clinical care could improve outcomes in patient undergoing treatment for hematologic malignancies and motivates future clinical trials to test this hypothesis. Validation of our findings using a larger cohort of patients with additional concurrent HFTM and SOC data is currently ongoing.

Flora:Medtronic: Current Employment. Ghosh:BMS: Consultancy; Cargo: Consultancy; Novartis: Research Funding; Kite/Gilead: Research Funding; Cabaletta Bio: Consultancy, Research Funding.