Performance of a multisensory implantable cardioverter-defibrillator algorithm for remote heart failure management: the RE-HEART registry

BACKGROUND: The HeartLogic algorithm measures data from multiple implantable cardioverter-defibrillator-based (ICD) sensors and combines them into a single index. The associated alert has proved to be a sensitive and timely predictor of impending heart failure (HF) decompensation. OBJECTIVE: To analyze the association between HeartLogic alerts and clinical events and to describe the implementation in clinical practice of the algorithm for remote management of HF patients. METHODS: The association between HeartLogic alerts and clinical events has been analyzed in the blinded study Phase 1 (from ICD implantation to HeartLogic alert activation) and in the following unblinded Phase 2 and 3 (after HeartLogic activation). During Phase 1, patients were managed according to the standard clinical practice and physicians were blinded to the alert status. During Phase 2 physicians reacted to alerts according to their clinical practice, while during Phase 3 they followed a standardized protocol in response to alerts. RESULTS: We enrolled 288 patients who received HeartLogic-enabled ICD or CRT-D at 15 centers. 101 patients contributed to Phase 1. During a median observation period of 10 [95% CI: 5 – 19] months, the HeartLogic index crossed the alert-threshold value 73 times (0.72 alerts/patient-year) in 39 patients. 8 HF hospitalizations and 2 emergency room admissions occurred in 9 patients (0.10 events/patient-year) during HeartLogic IN alert state. Other 10 minor events (HF in-office visits and non-HF hospitalization) were associated with HeartLogic alerts. During the active phases 267 patients were observed for a median follow-up of 16 [95% CI: 15 – 22] months. 277 HeartLogic alerts (0.89 alerts/patient-year) occurred in 136 patients. Thirty-three HeartLogic alerts were associated with hospitalizations for HF or with HF death (n=6), and 46 alerts were associated with unplanned in-office visits for HF. In 78 cases, HeartLogic alerts were associated with other clinically relevant events. The rate of unexplained alerts was low (0.39 alerts/patient-year). During the active phases, one HF hospitalization and one unplanned in-office visit for HF occurred when patients were in OUT of alert state. Patient phone contacts or in-person assessments were performed more frequently in Phase 3 (85%) than in Phase 2 (73%; p=0.047), while alert-triggered actions were similar in the two phases. Most alerts in both Phases (82% in 2 and 81% in 3; p=0.861) were managed remotely. The total number of patient phone contacts performed during Phase 2 was 35 (0.65 contacts/patient-year) and during Phase 3 was 287 (1.12 contacts/patient-year; p=0.002). CONCLUSIONS: HeartLogic index was frequently associated with HF-related clinical events, with a low rate of unexplained events. The HeartLogic alert and a standardize protocol of actions allowed to remotely detect impending decompensation events and to implement clinical actions with a low workload for the centers. FUNDING ACKNOWLEDGEMENT: Type of funding sources: None.