Remote monitoring of implantable electronic devices to predict heart failure decompensation

1

KEYWORDS:

• Cardiac implantable electrical device
• sensors
• device diagnostics
• heart
• failure
• remote monitoring

1. Introduction

The prevalence of heart failure (HF) is expected to increase by 46% from 2012 to 2030, placing a significant burden on healthcare resources. Healthcare expenditures for HF are primarily due to recurrent hospitalization [1], with a 30% rehospitalization rate within 60–90 days of discharge [2].

When acute episodes of HF (AHF) occur, the onset of symptoms is usually preceded by changes in several physiological parameters, taking place days before presentation to the emergency department. These variables can be monitored remotely, in the outpatient setting, thus enabling the detection of impending heart failure. In this context, the term remote monitoring (RM) encompasses both the telemedicine experience and the automatic transmission of parameters measured by implanted devices.

The former was based on patient-reported symptoms and signs, usually by means of phone calling, with high relevance of patient weight. More recently, the frequency of parameters recording was implemented with the use of apps and automatic transmission of these data. For instance, the EMPOWER trial evaluated an automated hovering intervention based on the use of an electronic pill bottle to measure adherence to daily medication and an electronic scale to measure daily weight [3].

The latter includes diagnostic devices implanted with the aim of monitoring specific parameters, as pulmonary artery pressure, in the case of CardioMEMS, and cardiac implantable electronic devices (CIEDs) that were implanted for pacing function or prevention of sudden cardiac death (including pacemaker, implantable cardioverter defibrillator, and cardiac resynchronization therapy).

These dedicated sensors have proven very effective for therapy optimization. Yet, their use is still very limited due to reimbursement and regulation issues. On the other hand CIEDs, have not been specifically to detect impending heart failure, but are currently implanted in a large share of HFrEF population for very consolidated indications, such as pacing, CRT, and sudden death primary prevention.

The aim of this review is to focus on the effectiveness and limitations of HF remote monitoring by means of implanted CIEDs.

2. Review

A large proportion of patients with heart failure have an implanted CIED, which can record relevant clinical information that could be used for heart failure detection and disease management.

This raised interest about the possibility that CIEDs can detect impending heart failure and the patient can be treated before overt decompensation occurs. CIEDs derived HF indicators are mostly nonspecific and unfit for automated AHF detection. Great expectations have been placed on thoracic impedance (TI), which is known to correlate with pulmonary capillary wedge pressure.

Unfortunately, TI reduction, considered as a single parameter, showed little predictive power and did not prove suitable for a strategy of early AHF detection and direct feed back to the patient. In fact, not only automatic alarm notification did not improve outcome, but it increased hospitalizations and outpatient visits and costs [4]. Similar results were obtained in the OptiLink HF study, comparing usual care versus delivering the notification alarm to the telemedicine staff instead of the patient; no benefit was found in the intervention arm (HR, 0.87; 95% CI: 0.72–1.04; P = 0.13). Furthermore, a large number of alarm state were not transmitted (24%) and only 30% of alarm condition required a medical intervention [5]

3. From device interrogation to remote monitoring

In parallel with the increase in diagnostic capabilities, major manufacturers of CIEDs developed platforms for the automated transmission of data collected by CIEDs. This made it possible to move from remote control to continuous remote monitoring (RM) of the devices. RM strategies for CIEDs yielded conflicting results. RM proved to be able to detect electrical problems of CIEDs and critical arrhythmias more promptly than conventional follow up [6]. Healthcare resources expenditure was reduced, mainly because of a lower number of in-office visits, but REM-HF and MORE-CARE trial failed to show a reduction of mortality or hospitalization [7,8]

3.1. Multi-parametric approach

A new approach to data interpretation was introduced by the PARTNERS-HF trial [9]. In this study, in-office device interrogations were retrospectively analyzed, searching for HF predictors. AF episodes > 6 hours, ventricular rate during AF > 90 bpm for more than 6 hours per day, OptiVol index > 60 were associated with AHF onset in the following 30 days. The presence of two or more of these conditions or a very high Optivol index (≥100) were associated to a 5.5-fold increase in risk of HF hospitalization within 30 days (hazard ratio: 4.8, 95% confidence interval: 2.9 to 8.1, p < 0.0001).

On this basis, using a Bayesian model and the variables of PARTNERS-HF, Cowie et al. developed a dynamic daily HF risk status score (HFRS) [10]. The highest risk score was associated with a 10-fold increased risk of HF hospitalization compared with a low-risk condition (adjusted HR: 10.0; 95% CI: 6.4–15.7, P < 0.001) [10].

The HFSR was validated by post hoc analyses in RAFT and MORE- CARE populations where it was able to identify individuals at high risk of hospitalization: OR 10.7 (95% CI: 6.9–16.6, P < 0.001) and 4.5 (95% CI: 3.1–6.6, P < 0.001), respectively.

4. Automatic identification of high HF risk condition in CIED patient

According to retrospective data analysis, multiparametric algorithms had been shown to be more efficient than single parameter assessment in HF risk assessment. Based on these observations, major manufacturers have been embedding similar algorithms in remote control systems (Figure 1).

Figure 1. Temporal development of HF risk assessment by means of remote monitoring tools.

Data flow is continuous, with at least daily transmissions that allow assessment of the patient’s current status and evaluation of its evolution over time

4.1. Medtronic experience: triage-HF

The Triage HF trial [11] enrolled 100 patients implanted with CIED, transmitting in case of predefined alert conditions, without a minimum transmission frequency. Whenever a high-risk status was detected, the subject would be contacted by phone within 24 hours. No contact was required for low-risk status and it was discretional for medium risk. The most frequent trigger condition was Optivol > 60 Ohms. Onset of HF worsening symptoms was reported in 63% of cases of high-risk status at HFSR score; the need for changing medical therapy was limited to 17% of the high-risk transmissions.

Triage HF real word application showed that integrating RM data with telephone triage in case of high-risk condition increases the sensitivity in the detection of acute HF episode (98.6%; 92.5–100.0%) while maintaining acceptable specificity (63.4%; 55.2–71.0%).

4.2. Boston scientific experience: HeartLogic^TM

The manufacturer Boston Scientific implemented its CIEDs with a specific algorithm for HF detection, named HeartLogic^TM, combining first and third heart sound (S1 and S3), S3/S1 ratio, thoracic impedance, respiration, night heart rate, and patient activity.

The HeartLogic^TM algorithm was validated in the MultiSENSE study [12]. The ‘In Alert’ state was associated to a high risk of decompensation (hazard ratio 24.53; CI 95, 8.55–70.38, P < 0.001). The sensitivity and specificity for AHF requiring hospitalization or intravenous diuretics were 70% and 87.5%, respectively. The ‘in-alert’ status had its onset 34–38 days before an acute HF event requiring hospitalization and 12 days before a minor HF event [12]. The unexplained alert rate was low (0.37–1.47 per patient year). The association of BNP to HeartLogic^TM index did not increase its predictivity [12].

Of note, preliminary data from an ongoing study showed that intervention after notification of ‘in-warning’ status was associated with fewer HF events (hazard ratio, 0.37 [95% CI, 0.14–0.99], P = 0.047). This result was a step up in the evaluation of remote monitoring efficacy because the algorithm not only was able to detect a pending HF episode but also permitted a specific intervention to avoid clinical HF episodes. It is possible that sensors based on heart sounds, having a hemodynamic value, improved the reliability of the algorithm.

4.3. Biotronik experience: the selene HF

The SeleneHF study recently analyzed the correlation between HF hospitalizations and deaths with Biotronik Home Monitoring™ data in patients implanted with ICD/CRT-D devices.

The hallmark of this algorithm is the introduction of the baseline Seattle Heart Failure Model score to stratify the risk assessed by the algorithm.

The study enrolled 918 patients, with a median follow up of 22.5 months, randomly assigned to the derivation or the validation group. The odds ratio for first HF hospitalization per unitary increase of index value was 2.73 (CI 1.98–3.78; P < 0.001) in the derivation group. In the validation group, sensitivity was 65.5% (CI 45.7–82.1%), and unexplained alert rate was 0.69 (CI 0.64–0.74) per patient-year [13].

Nowadays, the other manufacturers did not develop a multiparametric evaluation of sensors data.

5. Which are the evidence in favor of RM efficacy?

Retrospective studies agree in indicating that information obtainable from CIEDs, properly processed and transmitted to a telemedicine center, can be used to highlight situations with high risk of HF, even before symptoms have appeared.

Is this enough to say that RM is a useful tool for the management of patients with HF?

The unknowns are many. Is the sensitivity and specificity of the indicators/score sufficient? Are we able to intercept the process leading to AHF early enough? Are possible therapeutic interventions able to avoid the development of a state of overt cardiovascular decompensation, as suggested by MultiSENSE study? And is all this able to affect the patient’s prognosis or quality of life?

Actually, RM literature yields controversial outcomes. The main RM result is the reduction of time between the potentially harmful events (such as arrhythmias, system failures) and clinical intervention [6]. Thus, very high-risk patients are expected to benefit most from RM.

When it comes to harder endpoints such as hospitalization and mortality, results are inconsistent.

Only strictly standardized response protocols demonstrated the superiority of RM over usual care [6,8]. Of notice, none of the published prospective RCTs evaluates the multiparametric systems that are currently available and used, but refers to data processing systems that are, in fact, obsolete. None of the RCTs integrate the information acquired via RM and via telephone counseling, according to the common practice of most telemedicine centers.

‘Big-data’ analysis [14] suggests that regular use of first generation, non-multiparametric RM improves prognosis in HF patients. Actually, the benefit might be attributed to stricter clinical follow up protocols rather than to RM itself, so even this evidence is not clear.

Of note, the currently available randomized and non-randomized trials provide results that are highly dependent on the technologies employed and are difficult to apply to different telemedicine devices and platforms.

6. Challenges in RM adoption

The RM adoption has been slowed by logistic, legal, and economic issues.

In the RCTs the RM management includes, at least, the review of transmissions sent by patient and a telephone consultation upon which intervention is instated. A strictly organized workflow is required, where certain duties are assigned to every operator, including EP specialist, HF cardiologists, practitioners, nurses, in accordance to local practice and regulations. The workload depends on this organization and is roughly proportional to the number of enrolled patients.

Although, RM has been demonstrated cost effective from healthcare perspective [15], in many countries it is not or only marginally reimbursed by national Health System or insurance companies and the cost of RM lies on each individual health care provider (Hospital, CIEDs implanting center) having no interest in supporting a loss-making activity.

Privacy is a major issue both for transmission and storage of data which must comply with evolving standards and is perceived by medical operators as a possible source of risk.

7. Conclusions

In summary, RM-based follow-up strategies increase the timeliness of detection of various critical situations and reduce costs. Nevertheless, conclusive evidence of clinical benefit is lacking, with inconsistent effect on hospitalization and mortality. Further efforts are required to design and standardize response models for the management of HF patient by means of RM; meanwhile, healthcare systems are called to provide for an adequate reimbursement policy, according to benefits already demonstrated.

Declaration of Interest

GB Perego is a member of Medtronic European Advisory Board and received speakers’ fees from Medtronic, Abbott, & Boston Scientific. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Additional information

Funding

This paper was not funded.