Decongestive treatment adjustments in heart failure patients remotely monitored with a multiparametric implantable defibrillators algorithm

Abstract Aims HeartLogic algorithm combines data from multiple implantable defibrillators (ICD)‐based sensors to predict impending heart failure (HF) decompensation. A treatment protocol to manage algorithm alerts is not yet known, although decongestive treatment adjustments are the most frequent alert‐triggered actions reported in clinical practice. We describe the implementation of HeartLogic for remote monitoring of HF patients, and we evaluate the approach to diuretic dosing and timing of the intervention in patients with device alerts. Methods The algorithm was activated in 229 ICD patients at eight centers. The median follow‐up was 17 months (25th–75th percentile: 11–24). Remote data reviews and patient phone contacts were undertaken at the time of HeartLogic alerts, to assess the patient's status and to prevent HF worsening. We analyzed alert‐triggered augmented HF treatments, consisting of isolated increases in diuretics dosage. Results We reported 242 alerts (0.8 alerts/patient‐year) in 123 patients, 137 (56%) alerts triggered clinical actions to treat HF. The HeartLogic index decreased after the 56 actions consisting of diuretics increase. Specifically, alerts resolved more quickly when the increases in dosing of diuretics were early rather than late: 28 days versus 62 days, p < .001. The need of hospitalization for further treatments to resolve the alert condition was associated with higher HeartLogic index values on the day of the diuretics increase (odds ratio: 1.11, 95% CI: 1.02–1.20, p = .013) and with late interventions (odds ratio: 5.11, 95% CI: 1.09–24.48, p = .041). No complications were reported after drug adjustments. Conclusions Decongestive treatment adjustments triggered by alerts seem safe and effective. The early use of decongestive treatment and the use of high doses of diuretics seem to be associated with more favorable outcomes.


| INTRODUCTION
Heart failure (HF) is one of the leading causes of hospital admission worldwide, and is associated with high morbidity, mortality, and rehospitalization. 1 Remote monitoring and advanced diagnostic algorithms have been developed to provide continuous data on HF patients who receive implantable defibrillators (ICD) and resynchronization therapy (CRT-D). Many studies have reported combining ICD diagnostics to better stratify and manage patients at risk of HF events, 2,3 and current guidelines suggest that ICD-based multiparameter monitoring may be considered, to improve clinical outcomes. 1 The HeartLogic (Boston Scientific) algorithm combines data from multiple ICD-and CRT-D-based sensors and has proved to be a sensitive and timely predictor of impending HF decompensation. 4 Although its diagnostic performance in detecting HF worsening has been demonstrated, it is still not known what treatment protocol should be applied to manage the events notified by the algorithm.
Most of the symptoms associated with acute HF are the result of excessive fluid retention, and loop diuretics are the treatment of choice to combat them. 5 Decongestive treatment adjustments not only constitute the most common intervention when patients are hospitalized for acute HF, 6,7 but were also reported to be the most frequently triggered actions in response to alerts in the first experiences of HeartLogic use in clinical practice. 8,9 The aim of the present study was to describe the implementation of the HeartLogic algorithm in a protocol for the remote monitoring of HF patients, and to evaluate the approach to diuretic dosing and timing of the intervention in patients with device alerts.

| METHODS
The study was conducted in eight Italian high-volume arrhythmia centers. HeartLogic was activated in all HF patients with reduced left ventricular ejection fraction (≤35% at the time of implantation) who had received a HeartLogic-enabled ICD or CRT-D device (RESONATE family, Boston Scientific) between December 2017 and July 2020, in accordance with standard indications, 1 and were consecutively enrolled in the LATITUDE (Boston Scientific) remote monitoring platform. Patients were followed in accordance with the standard practice of the participating centers, based on current international recommendations. 10

| HeartLogic index
The details of the HeartLogic algorithm have been reported previously. 4 Briefly, the algorithm combines data from multiple sensors: accelerometer-based first and third heart sounds, intrathoracic impedance, respiration rate, the ratio of respiration rate to tidal volume, night heart rate, and patient activity. Each day, the device calculates the sensor-recorded values in terms of their shift from the baseline and computes a composite index. An alert is issued when this index crosses a programmable threshold. Weekly reminders (realerts) are sent until the HeartLogic index returns below the nominal alert recovery threshold. 5

| Alert management
The study protocol did not mandate any specific intervention algorithm, and physicians were free to remotely implement clinical actions (e.g., drug adjustments, educational interventions), to schedule extra in-office visits when deemed necessary for additional investigations or for interventions, or to adopt an active monitoring approach. In our analysis, we classified the alerts according to the management strategy adopted at the centers. We distinguished between alerts followed/not followed by clinical actions, and analyzed alert-triggered actions to treat HF (e.g., change in current HF medications, reinforcing adherence, device programming optimization). We then specifically investigated augmented HF treatments consisting of isolated increases in the equivalent dose of diuretics, as compared with the dose on the day before the initial alert.
Increases in the dosing of diuretics were categorized as either early or late, according to when they were initiated: early treatments were those implemented within 2 weeks of the first alert notification (following the initial alert or the first weekly reminder), while late treatments were those undertaken after 2 weeks (following the second or subsequent weekly reminders). Diuretic increases were also categorized as either major or minor. Administering >2 times the daily dose of loop diuretics or switching to a more bioavailable diuretic were considered major actions, 11 while lower increases were considered minor.
A shorter "in-alert" state duration was considered suggestive of the efficacy of the intervention and of its ability to resolve the alert condition without requiring further treatments.

| Statistical analysis
Descriptive statistics are reported as mean ± SD for normally distributed continuous variables, or medians with 25th to 75th percentiles in the case of nonnormal distributions. Normality of distribution was tested by means of the Kolmogorov-Smirnov test.
The time course of HeartLogic index and sensor changes surrounding the decongestive treatment adjustment were evaluated at four timepoints, as in a previous study. 12 A 30-day baseline was compared both with a 7-day preaction state measured up until the day before the diuretic augmentation, and with the state on the first day of the augmented therapy. Moreover, recovery was evaluated by recording sensor values over a 2-week period beginning 2 weeks after diuretic augmentation and comparing these with the baseline values. For control purposes, averaged sensor data were calculated in patients who did not have HF events and decongestive treatment adjustments during clinical follow-up. These trends were aligned on a random day during the observation period. Sensor data were compared between different temporal periods by means of a paired t-test. Differences in non-Gaussian variables were tested by means of the Mann-Whitney nonparametric test. Univariable binary logistic regression analysis was utilized to evaluate the relationship between the need for further treatment to resolve the alert condition and baseline clinical or treatment variables. All variables displaying a statistically significant difference (p < .05) were included in a multivariable binary logistic regression analysis. A p-value < .05 (two-tailed) was considered significant in all tests. All statistical analyses were performed by means of R: a language and environment for statistical computing (R Foundation for Statistical Computing).

| RESULTS
From December 2017 to July 2020, HeartLogic was activated in 229 patients who had received an ICD or CRT-D. Table 1 shows the baseline clinical variables of all patients. The median follow-up was 17 months (25th-75th percentile: 11-24) (a total of 308 patient-years).  amplitude and an increase in the respiratory rate and night heart rate.

| HeartLogic alerts and their management
Of the 56 decongestive treatment adjustments, 30 were implemented within 2 weeks of the first alert notification (early actions-average time from alert to intervention 5 ± 4 days), while the remaining 26 took place later (late actions-average time 40 ± 27 days). In 29 cases, a major increase in diuretic therapy was noted, while in 27 cases the increase in the daily dose was minor.  Average data from clinically stable periods are reported for comparison. The trends preceding the intervention were comparable between minor and major treatment adjustments. By contrast, before the action, the HeartLogic index was persistently higher in the case of late diuretic increases than early increases. In the subsequent period, the HeartLogic index decreased after decongestive treatment adjustments, and alert cases resolved more quickly when decongestive therapies were major and timely. The trends in the average sensed parameters that contribute to the calculation of the combined index are reported in Figures S1 and S2.

| Outcome of alert-triggered actions
Overall, timely diuretic changes were associated with a shorter "in- proved to be relevant and actionable, the rate of alerts judged nonclinically meaningful, and the rate of HF hospitalizations not associated with alerts being low. 8,9,14 Moreover, an alert-based management strategy seems more efficient than scheduled follow-up schemes.
T A B L E 2 Matched sensor data during baseline, preaction, first day of treatment, and recovery, stratified by time (early, late) and extent (major, minor) of intervention Note: Control group (clinically stable periods from patients with no HF events or decongestive treatment adjustments). Early (n = 30), late (n = 26), major (n = 29), minor (n = 27), control (n = 105).
Recent data have also suggested that activation of the multi-sensor algorithm might result in a significant reduction in hospitalizations for decompensated HF. 15 Recently, Calò et al. 9 analyzed the alert management strategies adopted at 22 centers and found that, when clinical actions were undertaken in response to alerts, the rate of HF events was lower. In Confirming previous findings, 12 we showed that, at the time of device detection of the HF event, the sensed parameters that contribute to the calculation of the HeartLogic index presented changes from their baseline reference values. In the present analysis, the accelerometerbased third heart sound, a surrogate of filling pressure, 16 significantly increased, as did the respiratory rate, which is known to facilitate the identification of patients at risk of worsening HF. 17 Similarly, night heart   25 The adoption of a multiparameter algorithm, that has proved to be sensitive and associated with a low rate of unexplained detections, 4 should overcome these limitations.

| Limitations
The main limitation of this study is its observational non-randomized design. Although the observational nature of the analysis may have introduced some biases, the consecutive enrollment should have decreased their magnitude. Indeed, the assignment of a HeartLogicenabled device was not guided by the characteristics of the patients.
Moreover, no predetermined actions were prescribed in response to HeartLogic alerts or to the individual subject's reported signs or symptoms. However, owing to the "real-world" nature of the study, this could even strengthen the usefulness of HeartLogic as a viable monitoring tool in everyday clinical practice.

| CONCLUSIONS
This study demonstrated the safety and efficacy of decongestive treatment adjustments triggered by HeartLogic alerts, even when such adjustments were completely dependent on the physicians' clinical expertise and were not standardized. Our results suggest that the early use of decongestive treatment and the use of high doses of diuretics are associated with more favorable outcomes.

DATA AVAILABILITY STATEMENT
The experimental data used to support the findings of this study are available from the corresponding author upon request.