Authors' objectives: Heart failure is a clinical syndrome caused by any structural or functional cardiac disorder that impairs the heart’s ability to function efficiently and pump blood around the body. Function can also be monitored using cardiac implantable electronic devices, some of which may also deliver a therapeutic benefit (e.g. pacemakers), while others only monitor metrics over time. Implantable devices can include algorithms that aim to predict the occurrence of a heart failure event. They are intended to be used alongside clinical judgement and make treatment decisions. To determine the clinical and cost-effectiveness of the four remote monitoring algorithms (CorVue, HeartInsight, HeartLogic and TriageHF) for detecting heart failure in people with cardiac implantable electronic devices. Heart failure (HF) is a clinical syndrome caused by any structural or functional cardiac disorder that impairs the heart’s ability to function efficiently and pump blood around the body. The most common symptoms of HF are breathlessness, fatigue and oedema. Conditions that can cause HF include coronary heart disease, high blood pressure, heart rhythm or valve abnormalities and conditions affecting the heart muscle (cardiomyopathies and myocarditis). Around 920,000 people in the UK were living with HF in 2018, with an estimated 200,000 new diagnoses each year. HF mainly affects people over the age of 65, with an average age of diagnosis of 77, and risk increases significantly with age. Around 1 in 35 people aged 65–74 years have HF, which increases to 1 in 15 of people aged 75–84 years, and to just over 1 in 7 people of those aged above 85 years. The National Institute for Health and Care Excellence (NICE) guidelines for diagnosis and management of chronic HF in adults recommend that monitoring of people with chronic HF should include a clinical assessment of functional capacity, fluid status, cardiac rhythm (minimum of examining the pulse), cognitive status and nutritional status, a review of medication and an assessment of renal function. The European Society of Cardiology (ESC) guidelines for the diagnosis and treatment of acute and chronic HF add that HF management may involve in-person service models or home-based telemonitoring, and that the COVID-19 pandemic has highlighted some of the potential advantages of the latter. While care is usually followed up by HF clinics, suitable patients may be followed up by community HF nurses or a general practitioner (GP) with a special interest in HF – a clinical expert commented that there is no standard HF service model. Patients who have cardiac implantable electronic devices (CIEDs) due to HF or who are at risk of HF may have a remote monitoring system incorporated in the device. The remote monitoring system includes a predictive algorithm for HF. The system can send alerts, and/or the stored data can be reviewed. There is additional cost to access and use the remote monitoring system. The decision question is whether the algorithm-based remote monitoring of HF risk data in people with CIEDs represent a clinical and cost-effective use of NHS resources and should be recommended for use. Four relevant remote monitoring algorithms were identified for consideration: CorVue algorithm with integrated CIED [Abbott Laboratories (Abbott Park, IL, USA)] HeartInsight algorithm with integrated CIED (BIOTRONIK, Berlin, Germany) HeartLogic algorithm with integrated CIED (Boston Scientific, Marlborough, MA, USA) TriageHF algorithm with integrated CIED [Medtronic plc (Medtronic plc, Dublin, Ireland)]. To determine the clinical and cost-effectiveness of the four remote monitoring algorithms for detecting HF in people with CIEDs.

Authors' results and conclusions: There was reasonable evidence to suggest HeartLogic and TriageHF can accurately predict heart failure events. CorVue’s prognostic accuracy is less clear due to high heterogeneity in findings between studies. There was only a single published HeartInsight study, which suggested similar accuracy to the other algorithms. Cost-effectiveness estimates could only be produced for HeartLogic and TriageHF, which were less costly and more effective compared to the respective cardiac implantable electronic device without the algorithms. For all technologies, only a small reduction in hospitalisation rates were required for them to be cost-effective. There was a lack of comparative evidence across all technologies included in the scope. Evidence for HeartLogic and TriageHF suggests that they may have acceptable prognostic accuracy for predicting heart failure events. However, further evidence is required to confirm these results. Specifically, further comparative evidence (e.g. randomised controlled trials) is required to show the benefit of the algorithms compared to standard practice in intermediate and clinical outcomes. For example, some studies suggested high false positive rates and low sensitivity. Only a single published study was identified for HeartInsight, therefore there are insufficient data to draw conclusions on prognostic accuracy and the benefits on clinical and intermediate outcomes. It is likely remote monitoring systems for CorVue, HeartInsight, HeartLogic and TriageHF would be cost-effective were they to result in fewer hospitalisations in heart failure patients; however, in general, this may apply to any device lowering the hospital visit. In addition, any potential benefits of reduced hospitalisation need to be carefully balanced with chances of overtreatment resulting from alerts. Clinical effectiveness Eighty-one reports comprising 42 studies of clinical effectiveness were included in the systematic review. Eight studies evaluated CorVue, 1 published study evaluated HeartInsight, 19 studies evaluated HeartLogic and 14 studies evaluated TriageHF. Of the included studies, the great majority were single-arm cohort designs (retrospective and prospective). No RCTs were identified, and five studies provided some comparative data (CorVue; n = 1, HeartLogic; n = 3, TriageHF; n = 1). The greatest amount of evidence for prognostic accuracy was identified in studies assessing the TriageHF algorithm (n = 10). Of these, the area under the curve (AUC) was reported in three studies assessing worsening HF (AUC= 0.75), mortality (AUC = 0.61) and hospital admissions (AUC = 0.81). Sensitivity for high risk status for HF-related events (e.g. hospitalisations) showed great variability (range = 37.4 to 87.9%). Specificity also varied (range = 44.4–90.2%). A similar amount of evidence was identified for prognostic accuracy outcomes evaluating HeartLogic (n = 8). Validation studies for HeartLogic to predict HF events (hospitalisations and clinical visits) reported sensitivity as adequate to high (range = 66–100%); similarly, specificity was adequate to high (range = 61–93%). False positive rates were generally low in the seven studies reporting this outcome; conversely, one study reported a high false positive rate (26 of 38 alerts). In comparison to HeartLogic and TriageHF, there was limited evidence for CorVue (n = 5) and HeartInsight (n = 1) overall and for prognostic outcomes. The CorVue algorithm demonstrated inadequate sensitivity for HF events, defined as hospitalisations (range = 20–68%). Specificity was only reported in two studies at 70% and 77%. The low predictive accuracy was also accompanied by a high false positive rate. The one published study for HeartInsight algorithm had 65.5% sensitivity and 86.7% specificity for first postimplant HF hospitalisations. Additionally, 54.8% sensitivity and 86.5% specificity for HF hospitalisation, outpatient intravenous intervention (IVI) or death. False positive rates were low. Reported clinical outcomes included HF events, mortality and adverse events (morbidity). Twelve studies reported HF events for three algorithms (HeartLogic, n = 7; CorVue n = 3, TriageHF n = 2). Only one of these studies was comparative, with data showing less HF events when the HeartLogic algorithm was utilised. For non-comparative evidence using HeartLogic, there was evidence that when IN alert compared to OUT of alert related to increased risk of HF events occurring. Similarly, TriageHF showed an increased risk of HF events when in high or medium risk status. No comparative evidence was generated for CorVue, and only numerical data were presented. No evidence was identified for HeartInsight. There was limited evidence for HF-related deaths. Three HeartLogic studies demonstrated an increased risk of death when IN alert compared to OUT of alert. One study assessing differences between unplanned HF hospitalisations and medical admissions for TriageHF reported more deaths occurring during HF hospitalisations. Only two studies reported adverse events (HeartInsight n = 1, HeartLogic n = 1). For the patient-reported outcome measures, one single prospective cohort evaluating the TriageHF algorithm provided outcomes for health-related quality of life by using the 6-minute walk test (6MWT) and Minnesota Living With Heart Failure Questionnaire (MLWHF). There was a decrease in walking distance at 8 months follow-up and no statistically significant change in the MLWHF from baseline to follow up at 8 months. The evidence for HeartLogic and TriageHF showed a potential to be of use in clinical practice; however, there are important uncertainties due to a lack of comparative evidence. HeartLogic had the highest and most consistent accuracy measures (i.e. sensitivity of ≥ 70%); the data suggest that being IN alert is linked to greater risk of HF events; however, these estimates were generally derived from composite outcomes (e.g. hospitalisations and outpatient visits). TriageHF showed similar accuracy, but with further degree of variation, for detecting such HF events when in a high risk status; however, these estimates were more commonly reported from single end-point studies. HeartInsight reported comparable accuracy to HeartLogic and TriageHF (sensitivity of 65%); however, this was only based on one published study, therefore it is uncertain whether further studies will replicate these findings. CorVue prognostic accuracy data varied substantially (i.e. sensitivity reported to be as low as 20%). For all technologies, most studies were judged to be at high risk of bias, which reduces certainty about the evidence. All remote monitoring algorithms only needed to reduce hospitalisations by a small amount for them to be cost-effective given the evidence on incremental healthcare visits use compared to no remote monitoring algorithm. Better quality and adequately powered evidence on both hospitalisations and healthcare contacts (visits, calls), which also records time spent reviewing remote monitoring data, would help inform the cost-effectiveness of the remote monitoring algorithms.

Authors' methods: We performed systematic reviews of clinical, cost-effectiveness, quality of life and cost outcomes. We searched MEDLINE and other sources of published and unpublished literature, including manufacturers’ websites and Clinical Trials Registries between June and August 2023. For the clinical effectiveness review, study selection was completed by two independent reviewers at both title and abstract, and full-text screening stages. Data extraction and study quality appraisal were completed by a single reviewer and checked for accuracy by a second. Due to heterogeneity, no statistical analyses were performed, and a narrative synthesis was reported. A de novo two-state Markov model (with alive and dead states) was used to estimate the cost-effectiveness of algorithm-based remote monitoring of heart failure risk data in people with cardiac implantable electronic devices over a lifetime. The evidence for each algorithm was limited in terms of comparative evidence. Additionally, available evidence was often of low quality. The comparative outcome evidence for economic model was very limited. Systematic review The systematic review was conducted following the general principles recommended by the Centre for Reviews and Dissemination (CRD) guidance. A comprehensive range of databases and sources of grey literature were searched for the identification of studies relating to the use of algorithm-based remote monitoring of HF risk data in people with CorVue, HeartInsight, HeartLogic or TriageHF CIEDs. The bibliographic databases searched were MEDLINE and EMBASE (via Ovid), the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO), the Cochrane Database of Systematic Reviews (CDSR) and the Cochrane Central Register of Controlled Trials (CCRCT) (via the Cochrane Library) and the Database of Abstracts of Reviews of Effects (DARE) (via the CRD). International CTRIs, such as the US ClinicalTrials.gov; EudraCT; the World Health Organization’s International Clinical Trials Registry Platform (ICTRP); and ScanMedicine, a multinational open access clinical trial database, were searched for the identification of ongoing clinical trials. Additionally, we searched for health technology assessment (HTA) reports in the international HTA database (INAHTA) and for protocols of systematic reviews in International Prospective Register of Systematic Reviews (PROSPERO) and International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY), both international registers of systematic reviews. Finally, we searched for pre-print manuscripts in medRxiv, a pre-print server for health sciences. We performed backwards and forwards citation chaining to identify potentially relevant studies cited or citing the included studies. Company submission documents and company websites were also searched for additional relevant studies. Data extraction of the study characteristics and outcome data was done by one reviewer and checked by another reviewer. The risk of bias was assessed using Risk Of Bias In Non-randomized Studies – of Interventions (ROBINS–I), whereby clinical outcomes were reported in non-randomised intervention studies; and Prediction model Risk Of Bias Assessment Tool (PROBAST), whereby prognostic outcomes, including sensitivity and specificity, were reported. The Cochrane risk of bias tool was not used because none of the included studies were randomised controlled trials (RCTs). Due to diversity across the studies, meta-analysis was not performed, and the evidence was synthesised narratively and in tabular format.

Project Status: Completed

URL for project: https://www.journalslibrary.nihr.ac.uk/programmes/hta/NIHR135894

Year Published: 2025

URL for published report: https://www.journalslibrary.nihr.ac.uk/hta/PPOH2916

URL for additional information: English

English language abstract: An English language summary is available

Publication Type: Full HTA

Country: England, United Kingdom

DOI: 10.3310/PPOH2916

Organisation Name: NIHR Health Technology Assessment programme

Contact Address: NIHR Journals Library, National Institute for Health and Care Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK

Contact Name: journals.library@nihr.ac.uk

Contact Email: journals.library@nihr.ac.uk