Telemonitoring with an implantable loop recorder in outpatient heart failure care

This study was a prospective, observational multicentre study performed in three large Dutch non-academic teaching hospitals (trial registration: NCT01366703). Patients were recruited between July 2011 and April 2014 and were included after giving written informed consent. Minimal follow-up duration was 1 year. Because of the nature of the study (descriptive without statistical assumption because of lack of previous data) no formal power calculation was done to determine the sample size. A patient sample of 30 was considered reasonable. The study was approved by the Medical Research Ethics Committee of Noord-Holland.

Patients with stable systolic heart failure (ejection fraction 35–45%) or diastolic heart failure (ejection fraction > 45% and echocardiographic signs of impaired diastology or NT-proBNP >250 ng/l) were included. Stable was defined as a New York Heart Association (NYHA) Class II or III, without hospitalisation in the last 3 months or a scheduled intervention in the coming 3 months. Patients with a pacemaker/ICD or an indication for such a device were excluded. Since we sought possible therapeutic implications from continuous monitoring, patients with known AF, a virtual CHA₂DS₂-VASc of 1 or less or already on an OAC were excluded.

After inclusion, the implantable cardiac monitor Reveal XT (Medtronic Inc., Minneapolis, USA) was inserted in a routine fashion. Patients were placed on home monitoring with Carelink and the following alerts activated: ventricular tachycardia, asystole of longer than 3 seconds, bradycardia and AF. All Carelink alerts were checked daily by a device technician during in office hours on weekdays.

Patients were followed in a parallel fashion. First, they were seen at the clinic by a research nurse, experienced in heart failure trials, at intake, 6 and 12 months, with additional structured telephonic interviews at month 3 and 9. The research nurse was also responsible for careful case report form (CRF) compliance and CRF documentation, collection of the pre-specified arrhythmic events in the CRF and distribution of the pre-specified arrhythmic events to the treating physician (see later). The ILR was interrogated before visiting the research nurse at month 6 and 12. Second, in parallel, patients were seen by their own treating physician the same day at month 6 and 12. Minimal follow-up was 1 year. Their treating physician was unaware of the ILR findings except for several pre-specified events. Pre-specified events were asystolic episodes of 3 seconds or more, AF or other supraventricular arrhythmia longer than 6 minutes, non-sustained ventricular tachycardia or second/third degree AV block of any duration. If a pre-specified event occurred (detected at a routine visit to the technician or by a home-monitoring alert), the device technician contacted the research nurse who then took appropriate action (CRF update and informing the treating cardiologist). Therapeutic interventions (whether based on a clinical event or after a pre-specified event from the ILR) were made at the discretion of the treating physician.

All data were processed using IBM SPSS 20.0. Continuous data are expressed as mean ± standard deviation, discrete data are expressed as frequency and percentage.

Data retrieved from Carelink were additionally processed at Medtronic Bakken Research (Heerlen; the Netherlands) to obtain aggregated data on arrhythmic events and AF burden, which were exported to an excel sheet. All event tracings were assessed by an experienced device cardiologist (RT).

Baseline characteristics are presented in Tab. 1. Between July 2011 and April 2014, 30 patients were enrolled and received an ILR. The mean age was 71.8 ± 8.8 years. The majority were female (56.7%). At baseline, 25 patients were in NYHA functional class II and 5 in class III. The mean ejection fraction was 44.5 ± 7.8 and 80% of the patients had systolic heart failure.

Median follow-up was 12 months. No patients were lost to follow-up. Two ILRs were removed, at 2 and 12 months, due to local discomfort. One patient had a superficial wound infection. Three patients died during follow-up. The first patient with severe triple vessel disease and inoperable left main coronary stenosis died of acute cardiogenic shock within 1 month after ILR implantation. During this month, no events were recorded by the ILR. The second patient died due to a ruptured abdominal aortic aneurysm within 1 month after inclusion, the ILR data were lost. The last patient died in his sleep, no rhythm or conduction disorders were recorded by the ILR. No patients were admitted to the hospital for worsening heart failure.

In 15 patients (50%) pre-specified events were detected (Tab. 2). The mean time to a pre-specified event was 6.0 ± 3.6 months (Fig. 1). AF was recorded in 8 patients (26.7%). ILR data led to therapeutic changes in 13 patients (43.3%). The following therapeutic changes were implemented: pacemaker implant in 1, institution of OAC in 8 (vitamin K antagonists in 7, direct-acting oral anticoagulants in 1) and adjustments of beta-blocker dose in 6 patients. Of the pre-specified ILR events, none was detected by the routine clinic visits to the cardiologist.

In 20 patients AF was detected by the ILR, 12 were considered false-positive after expert adjudication. Therefore 8 patients were diagnosed with true AF of >6 min. By increasing the minimal duration of AF detection to >60 min, AF was detected in 13 patients by the ILR, 6 were labelled false-positive after expert adjudication leaving 7 patients with true AF of >60 min. The mean AF burden was 65.8 ± 173.2 min/day.

Firstly, therapy change was left to clinical judgment in our study. The institution of OACs in case of subclinical AF detected by cardiac implantable devices is under debate [19]. The impact of OAC therapy (including direct anticoagulants) in subclinical device-detected AF is not yet proven and currently being studied in the NOAH-AF trial (NCT02618577).

And second, the study was not randomised but to a certain extent the patients served as their own control group, because treating physicians were mainly blinded to the ILR findings. Although pre-specified findings were reported to the cardiologist, none of these were detected by the routine visits of standard care. Physicians might have been biased knowing that they would be provided with the ILR findings should these occur. Also, this finding does not necessarily imply that the pre-specified events would not be detected clinically later on, but telemonitoring has proven to detect clinically relevant events much earlier. In the Trust trial, the median time to evaluation was <2 days in the home monitoring group compared with 36 days in patients without home monitoring [20]. These findings were reproduced in the Connect trial were the median time from a clinical event to clinical decision per patient was reduced from 22 days in the in-office arm without home monitoring to 4.6 days in patient on remote monitoring [21].

Finally, only 30 patients were included over several years. No one reason can be given for this slow inclusion; however, precise data are missing since no screening log was kept. During the trial inclusion period, several competing heart failure trials were running in our departments. Also, some suitable candidates were unwilling to undergo an ILR implant.