Telemonitoring for heart failure: a meta-analysis

Abstract

Aims

Telemonitoring modalities in heart failure (HF) have been proposed as being essential for future organization and transition of HF care, however, efficacy has not been proven. A comprehensive meta-analysis of studies on home telemonitoring systems (hTMS) in HF and the effect on clinical outcomes are provided.

Methods and results

A systematic literature search was performed in four bibliographic databases, including randomized trials and observational studies that were published during January 1996–July 2022. A random-effects meta-analysis was carried out comparing hTMS with standard of care. All-cause mortality, first HF hospitalization, and total HF hospitalizations were evaluated as study endpoints. Sixty-five non-invasive hTMS studies and 27 invasive hTMS studies enrolled 36 549 HF patients, with a mean follow-up of 11.5 months. In patients using hTMS compared with standard of care, a significant 16% reduction in all-cause mortality was observed [pooled odds ratio (OR): 0.84, 95% confidence interval (CI): 0.77–0.93, I²: 24%], as well as a significant 19% reduction in first HF hospitalization (OR: 0.81, 95% CI 0.74–0.88, I²: 22%) and a 15% reduction in total HF hospitalizations (pooled incidence rate ratio: 0.85, 95% CI 0.76–0.96, I²: 70%).

Conclusion

These results are an advocacy for the use of hTMS in HF patients to reduce all-cause mortality and HF-related hospitalizations. Still, the methods of hTMS remain diverse, so future research should strive to standardize modes of effective hTMS.

Structured Graphical Abstract

Summary results for all-cause mortality, first heart failure hospitalization, and total/recurrent heart failure hospitalizations divided in invasive home telemonitoring systems and non-invasive home telemonitoring systems and total. I² represents heterogeneity between studies. CI, confidence interval.

Telemonitoring, Heart failure, Non-invasive, Invasive, Mortality, Hospitalization

See the editorial comment for this article ‘Is telemonitoring for heart failure ready after a journey longer than two decades?’, by F. Koehler and G. Hindricks, https://doi.org/10.1093/eurheartj/ehad395.

Introduction

Heart failure (HF) is a chronic, complex, and progressive syndrome with a significant impact on public health. Globally, >60 million patients are affected by HF, and with the ageing of the general population, its prevalence is expected to increase in the forthcoming years.¹ Despite advances in medical therapy, cardiac implantable electronic devices (CIEDs) and (long-term) mechanical circulatory support, the morbidity and mortality of HF remain high. Moreover, HF places a high burden on healthcare due to frequent outpatient follow-up and recurrent hospitalizations as a result of deterioration of HF.² The costs of HF care are projected to further increase, primarily driven by hospitalizations. Therefore, there is a great need to develop effective strategies to reduce HF (re-)admissions and improve ambulatory HF care. Telemonitoring by means of home telemonitoring systems (hTMS) in this respect seems a promising option, which has gained even more momentum after the COVID-19 pandemic.³ The hTMS is a system at home, which uses a non-invasive or invasive device to collect health data, such as vital signs and other diagnostic data.⁴ While the number of studies—both randomized controlled trials (RCTs) and observational—reporting on hTMS has increased rapidly over the last years, their results and applicability have been uncertain due to heterogeneity. ^5–8

In 2015, a comprehensive Cochrane meta-analysis demonstrated a minor, albeit statistically significant reduction in all-cause mortality (ACM) through the use of structured telephone support (STS) and a significant reduction of both HF hospitalizations (HFH) and ACM by employing other non-invasive telemonitoring solutions.⁵ However, the results are hampered by high heterogeneity between the individual studies due to the differences in methodology of the employed systems, some risk of bias, and the lack of a consistent effect in many studies individually. Furthermore, there is a lack of studies pertaining to real-world data and repeated events in this meta-analysis. This conflicting evidence has led to a weak (class IIb, LoE B) recommendation for hTMS in the latest ESC Guidelines on Acute and Chronic HF.¹

However, medical technology is ever evolving, and newer hTMS have been developed including invasive devices such as CIEDs incorporating new algorithms to detect deterioration of HF (e.g. Heartlogic) and invasive haemodynamic devices measuring the pulmonary artery pressure (e.g. CardioMEMS and Cordella). Also, non-invasive remote monitoring strategies have improved and are now more structured, like the system used in the TIM-HF2 trial.⁹ Moreover, the COVID-19 pandemic has further accelerated the process of employing hTMS within the context of HF management.³ In order to fill in the abovementioned knowledge gaps, it is of great importance to explore the ever-growing body of contemporary literature regarding this subject as the HF community is on the breach of an outbreak of telemonitoring integration in clinical practice. Therefore, we performed a systematic review and meta-analysis of both RCTs and observational studies up to July 2022, comparing hTMS with standard of care (SoC) in patients with HF and describe the efficacy on clinical endpoints.

Methods

Protocol and registration

We performed a systematic review and meta-analysis of prospective studies (RCTs and observational studies) following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.¹⁰ This review is registered in the International Prospective Register of Systematic Reviews with number CRD42022306677.

Search strategy and selection criteria

In collaboration with an expert librarian specialized in systematic searches, a literature search was carried out, including studies that were reported during 1996—1 July 2022, using Embase, Medline Ovid, Web of Science, and Cochrane CENTRAL. Keywords used in the search were ‘heart failure’, ‘telemonitoring’, ‘implantable haemodynamic monitor’, ‘implantable cardioverter defibrillator’, ‘home monitoring’, ‘e-health’, ‘clinical trial’, and ‘prospective study’. The full search strategy is presented in the Supplementary material. Only published peer-reviewed original articles in the English language were included in our study. In addition, cross-referencing for any additional eligible studies was performed.

Studies were included if they contained any form of hTMS in chronic HF patients aged 18 years or older. We defined hTMS as a system in the home setting that employs a non-invasive or invasive device to remotely collect vital signs and other biometric or health-status related data (such as weight, blood pressure, heart rate, pulmonary pressure, ECG lead, and signs and symptoms with the exception of physical activity) and remotely transmits the collected data to a healthcare institution for further assessment by a healthcare provider. All other eligibility criteria are presented in Supplementary data online, Table S1.

Three reviewers (N.S., M.G., and D.A.) independently performed title and/or abstract screening in order to identify studies that potentially met the inclusion criteria. Results were then discussed, and any disagreement regarding eligibility was resolved by consensus. The full text of these studies was then retrieved and read independently by the same reviewers. Hereafter, each study was discussed in detail to decide upon the eligibility based on the inclusion and exclusion criteria. In case no consensus was reached, the principal investigator (J.B.) had the final say.

If eligible studies described the same population, only the study with the longest follow-up or most recent publication (with an active intervention arm) containing the entire population was included, unless different outcomes of interest were studied in each article. Studies describing a subgroup of the same population were excluded.

Data extraction, home telemonitoring systems categories, and study endpoints

The following information was extracted from the main study reports: author, year of publication, country, study name, study design, enrollment years, sample size, age, sex, New York Heart Association (NYHA) class, left ventricular ejection fraction (LVEF) cut-off, HF aetiology, medication use, type of telemonitoring solution, comparison group, follow-up duration, and endpoints. If studies presented endpoints at more than one time point, endpoints from the latest time point were extracted. Data extraction was performed by M.G., N.S., and D.A., independently. Categories and definitions of non-invasive and invasive hTMS and subcategories are presented in Table 1. Non-invasive hTMS consisted of the following separate subcategories: telemonitoring (TM), structured telephone support (STS), and a combination of TM and STS (complex TM). Invasive hTMS consisted of the following separate subcategories: cardiac implantable electronic devices (CIED), and invasive haemodynamic monitoring (IHM) (Table 1). The primary outcomes for this meta-analysis were ACM, first HFH, and total number of HFHs.

Table 1

Categories of telemonitoring modalities and abbreviations used

Home Telemonitoring System		Definitions
Non-invasive hTMS
– TM	Telemonitoring (individual)	Modality in which biometric data and/or health-related questionnaires are collected and sent to an HF clinic.
– STS	Structural telephone support	Modality in which HF patients are called by a HF nurse or cardiologist on a frequent basis.
– Complex TM	Complex telemonitoring	Modality in which multiple TM is combined with STS and/or 24-h call center or mix of other sub-modalities.
Invasive hTMS
– CIED	Cardiac implantable electronic devices	Modality in which PM/ICD systems (optionally with impedance leads) are used to monitor the patient.
– IHM	Invasive haemodynamic monitoring	Modality in which invasive haemodynamic parameters are used, e.g. (pressure) sensors.

hTMS, home telemonitoring system; TM, telemonitoring; STS, structural telephone support; HF, heart failure; CIED, cardiac implantable electronic device; PM, pacemaker; ICD, implantable cardiac defibrillator; IHM, invasive haemodynamic monitoring.

Quality assessment

The Cochrane risk of bias (RoB2) and ROBINS-I were used to assess the risk of bias for RCTs and observational studies, respectively. Each article was assessed independently by at least two authors (N.S., M.G., and/or D.A.). In case no consensus was reached, a third author was available for consultation to give their conclusive opinion.

Statistical analysis

Continuous variables are presented as means and ± standard deviations (SD) or medians and interquartile ranges (IQR), as appropriate. Categorical variables are presented as counts and percentages. Random-effects methods were used to obtain an estimate of the pooled treatment effect, applying the DerSimonian and Laird procedure. For ACM and first HFH, we present the pooled treatment effect as odds ratio (OR) and corresponding 95% confidence interval (CI). The endpoint total HFHs is presented as incidence rate ratio (IRR) and 95% CI, which required person-years to be calculated. Person-years were calculated by using the mean or median follow-up time. If no mean or median follow-up time was available, the planned follow-up time was used, with the exception of patients who withdrew or died. To calculate person-years for these patients, we used half of the planned follow-up time.

Sensitivity analyses were performed based on the specified categories of hTMS (Table 1). In this meta-analysis, we use ORs as the key effect measure, since the data that are required to obtain this measure can be directly derived from the study reports. It is true that in scenarios where the time varies, the hazard ratio (HR) is the preferred effect measure. However, HRs are only presented in 30% (endpoint hospitalization) to 37% (endpoint mortality) of studies, and must thus be estimated for the other studies. Therefore, we decided to present the (pooled) HRs as a sensitivity analysis and not as main analysis. Furthermore, another sensitivity analysis was carried out, dividing studies in short- (<3 months), mid- (3 to 12 months) and long-term (>12 months) follow-up time with respect to ACM and first HFH. Heterogeneity was assessed using the I²-statistic and classified as not important (I²: ≤25%), moderate, (I²: 26%–50%), substantial (I²: 51%–75%), and considerable (I²: >75%).¹¹ Funnel plots were generated and Egger regression tests performed to assess publication bias. All analyses were carried out using R Studio version 3.0 with the Metafor 3.4–0 package. A two-sided P-value of ≤0.05 was considered as statistically significant.

Results

Study characteristics

The literature search exposed, after duplicate removal, 6112 studies. A total of 91 studies that met all the eligibility criteria were included. In addition, one study was added from cross-referencing, resulting in a total of 92 studies.^9,12–102 The full PRISMA-flow diagram is shown in Figure 1. Within the 92 included studies, 36 549 HF patients were included, with a mean follow-up of 11.5 (range: 1.0–34.9) months. A total of 23 610 HF patients were included from 65 non-invasive hTMS studies, with a mean follow-up of 9.9 (range: 1.0–32.4) months (Table 2; Supplementary data online, Table S2). In 27 invasive hTMS studies, 12 939 HF patients were included and had a mean follow-up of 15.3 (range: 5.4–34.9) months (Table 3; Supplementary data online, Table S3). In non-invasive hTMS and invasive hTMS, 8 and 11 studies, respectively, were observational (either pre–post studies, matched studies, or single arm studies). All other studies were RCTs.

Figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow-chart.¹⁰³

Table 2

Trial characteristics non-invasive studies

Author, year (study)	Country	Design	Enrollment	n	Age, years	Men, %	NYHA	LVEF	Ischaemic aetiology, %
							III–IV, %	cut-off
Angermann et al., 2012 (INH)15	DE	RCT	2004–2007	715	68.6 ± 12.2	71	40	ANY	58
Antonicelli et al., 200816	IT	RCT	NA	57	78 ± 8.5	61	42	ANY	67
Baker et al., 201117	US	RCT	2007–2009	605	60.7 ± 13.1	52	31	ANY	NA
Balk et al., 200818	NL	RCT	2005–2006	214	66 (33–87)a	70	52	ANY	57
Bento et al., 200919	BR	RCT	NA	40	57.5 ± 9.4	70	38	ANY	25
Blum et al., 2014 (MCCD)20	US	RCT	2001–2005	203	72.5 ± 9	71	86	ANY	65
Boyne et al., 2012 (TEHAF)13	NL	RCT	2007–2008	382	71.4 ± 11.2	59	43	ANY	50
Capomolla et al., 200425	IT	RCT	2000–2001	133	57 ± 10	88	33	ANY	41
Chaudhry et al., 2010 (Tele-HF)26	US	RCT	NA	1653	61 (51–73)	58	57	ANY	51
Chen et al., 201027	TW	NRCT	2003–2005	550	68.2 ± 15.5	71	NA	<45%	58
Cichosz et al., 2018 (Danish telecare north)29	DK	RCT	NA	299	70.5	81	NA	ANY	NA
Cleland et al., 2005 (TEN-HMS)30	NL/UK/DE	RCT	2000–2002	426	67.2 ± 11.6	77	34	<40%	78
Copeland et al., 201032	US	RCT	2005	458	70.0 ± 10.8	99	44	ANY	35
Comin-Colet et al., 2016 (iCOR)31	ES	RCT	2010–2012	178	74 ± 11	59	54	ANY	35
Dar et al., 2009 (Home HF)34	UK	RCT	2006–2007	182	71.0 ± 11.7	66	NA	ANY	55
De Lusignan et al., (2001)35	UK	RCT	NA	20	75.2	NA	NA	ANY	NA
DeBusk et al., 200437	US	RCT	1998–2001	462	72 ± 11	51	50	ANY	51
Delaney et al., 201338	US	RCT	2011–2012	100	NA	32	100	ANY	NA
Dendale et al., 2012 (TEMA-HF)39	BE	RCT	2008–2010	160	75.8 ± 9.7	65	NA	ANY	NA
DeWalt et al., 200640	US	RCT	2001–2003	127	62.5 ± 10.1	49	50	ANY	NA
Domingues et al., 201042	BR	RCT	2005–2008	120	63 ± 13	58	NA	<45%	NA
Galbreath et al., 200443	US	RCT	1999–2003	1069	70.9 ± 10.3	71	24	ANY	NA
Galinier et al., 2020 (OSICAT)44	FR	RCT	2013–2016	990	70 ± 12.4	72	49	ANY	NA
Gambetta et al., 200745	US	NRCT	NA	282	74.6 ± 13	56	NA	ANY	46
Gattis et al., 1999 (PHARM)46	US	RCT	1996–1997	181	NA	68	33	<45%	NA
GESICA, Grancelli et al., 2005 (DIAL)	AR	RCT	2000–2001	1518	65 ± 13.3	71	49	ANY	NA
Giordano et al., 200947	IT	RCT	2002–2004	460	57 ± 10	85	40	<40%	53
Gjeka et al., 202148	US	RCT	2016–2018	62	68.6	49	NA	NA	NA
Goldberg et al., 2003 (WHARF)49	US	RCT	1998–2000	280	59.1 ± 15.3	68	100	≤35%	43
Ho et al., 200752	TW	OBS (pre-post)	2004	247	60 ± 17	68	33	≤40%	49
Kalter-Leibovici et al., 201755	IL	RCT	2007–2012	1360	70.7 ± 11.3	73	85	ANY	NA
Kashem et al., 200856	US	RCT	NA	48	53.7 ± 10.5	74	58	ANY	41
Köberich et al., 201557	DE	RCT	2011–2013	110	61.7 ± 12.0	83	34	≤40%	53
Koehler et al., 2011 (TIM-HF)58	DE	RCT	2008–2009	710	66.9 ± 10.6	81	50	≤35%	56
Koehler et al., 2018 (TIM-HF2)9	DE	RCT	2013–2017	1538	70.0 ± 10.5	70	48	ANY	41
Kotooka et al., 2018 (HOMES-HF)59	JP	RCT	2012–2013	181	66.2 ± 14.2	59	22	ANY	30
Krum et al., 2013 (CHAT)60	AU	RCT	2003 -?	405	73.0 ± 10.5	63	41	<40%	NA
Laramee et al., 200363	US	RCT	1999–2001	287	70.7 ± 11.8	54	36	<40%	71
Lyngå et al., 2012 (WISH)67	SE	RCT	NA	319	73.6 ± 10.1	75	100	<50%	46
Mo et al., 202168	CN	OBS	2019	300	53.1 ± 11.4	67	52	<40%	NA
Morguet et al., 200870	DE	OBS (matched)	2004–2006	128	60.8 ± 10.2	88	25	≤60%	69
Mortara et al., 2009 (HHH)71	UK/IT/PL	RCT	2002–2004	461	60 ± 12	85	40	≤40%	56
Negarandeh et al., 201973	IR	RCT	2016	80	NA	60	NA	NA	NA
Nouryan et al., 201974	US	RCT	NA	89	83.2	32	NA	NA	NA
Nunes-Ferreira et al., 202075	PT	OBS (matched)	2016–2018	125	65.9 ± 11.9	68	8	≤40%	38
Olivari et al., 2018 (RENEWING HEALTH)76	EU	RCT	2011–2014	339	80.0 ± 7.0	63	52	ANY	43
Ong et al., 2016 (BEAT-HF)77	US	RCT	2011–2013	1437	73	54	75	ANY	NA
Pedone et al., 201578	IT	RCT	NA	96	80 ± 7	39	68	ANY	NA
Pekmezaris et al., 201979	US	RCT	2014–2016	104	59.9 ± 15.1	57	70	ANY	NA
Pérez-Rodríguez et al., 2015	MX	RCT	2011–2012	40	68.2 ± 7.5	65	100	NA	NA
Ramachandran et al., 200780	IN	RCT	2005	50	44.6 ± 11.9	78	26	<40%	12
Riegel et al., 200281	US	RCT	NA	358	73.8 ± 12.4	51	97	ANY	49
Ritchie et al., 201682	US	RCT	2010–2011	346	63.3 ± 13.1	51	NA	NA	NA
Roth et al., 200484	IL	OBS	NA	118	74 ± 9	70	78	<50%	NA
Scherr et al., 200986	AU	RCT	2003–2008	120	NA	73	87	NA	NA
Schwarz et al., 200887	US	RCT	NA	102	78.1 ± 7.1	48	79	ANY	NA
Seto et al., 201288	CA	RCT	2009–2010	100	53.7 ± 13.7	79	46	<40%	33
Soran et al., 2008 (HFHC trial)92	US	RCT	2002–2005	315	76 ± 7	35	42	≤40%	55
Villani et al., 201496	IT	RCT	NA	80	72 ± 3	74	NA	<40%	NA
Völler, et al., 202297	DE	RCT	2010–2013	621	63.0 ± 11.5	88	31	<40%	59
Vuorinen et al., 2014 (Heart at Home)98	FI	RCT	2010–2012	94	58.1 ± 11.8	83	62	≤35%	NA
Wagenaar et al., 2019 (e-VITA HF)99	NL	RCT	2013–2014	450	66.8 ± 11.0	74	20	NA	NA
Wakefield et al., 2008100	US	RCT	2002–2005	148	69.3 ± 9.6	99	72	NA	NA
Ware et al., 2020101	CA	OBS (pre-post)	2016–2019	315	58.3 ± 15.5	78	31	ANY	NA
Wita et al., 2022102	PL	RCT	2014–2017	63	66.1 ± 10.5	87	NA	NA	29

SD, standard deviation; IQR, interquartile range; NYHA, New York Heart Association classification; LVEF, left ventricular ejection fraction; DE, Germany; IT, Italy; US, United States; NL, The Netherlands; BR, Brazil; TW, Taiwan; DK, Denmark; UK, United Kingdom; ES, Spain; BE, Belgium; FR, France; AR, Argentina; IL, Israel; JP, Japan; AU. Australia; SE, Sweden; CN, China; PL, Poland; IR, Iran; PT, Portugal; EU, Europe; MX, Mexico; IN, India; CA, Canada; TH, Thailand; FI, Finland; RCT, randomized controlled trial; OBS, observational study; NRCT, non-randomized controlled trial; NA, not available.

Median (range).

Table 3

Trial characteristics invasive studies

Author, year (study)	Country	Design	Enrollment	n	Age, years	Men %	NYHA III–IV, %	LVEF cut-off	Ischaemic aetiology, %
Abraham et al., 2016 (CHAMPION)12	US	RCT	2007–2009	550	61.6 ± 12.8	73	100a	ANY	61
Adamson et al., 2011 (REDUCEhf)13	US	RCT	NA	400	55 ± 15	69	51	ANY	45
Angermann et al., 2020 (MEMS-HF)14	NL/DE/IE	NRCT	2016–2018	234	67.9 ± 10.7	78	100	ANY	53
Böhm et al., 201621	DE	RCT	2008–2013	1002	66.3 ± 10.4	80	87a	<35%	54
Boriani et al., 2016 (MORE-CARE)	IT	RCT	2009–2014	918	66 ± 10	76	62	ANY	44
Bourge et al., 2008 (COMPASS-HF)23	US	RCT	NA	274	58 ± 13.5	65	100	<50%	81
Chiu et al., 2021 (REMOTE-CIED)	NL/DK	RCT	2013–2016	595	65 (59–73)	78	33a	ANY	55
Cowie et al., 2022 (COAST)33	UK	OBS	2017–2018	100	69 ± 11.9	70	100	ANY	39
De Simone et al., 2015 (EFFECT)36	IT	NRCT	2011–2013	987	66 ± 12.5	77	44	ANY	55
Domenichini et al., 2015 (LIMIT-CHF)41	UK	RCT	2010–2013	80	67.9 ± 11.4	94	NA	<50%	NA
Hansen et al., 2018 (InContact)50	DE	RCT	2010–2014	210	63.8 ± 11.1	84	43	≤35%	59
Hindricks et al., 2014 (IN-TIME)51	AU/EU/IL	RCT	2007–2010	664	65.5 ± 9.4	81	57	≤35%	NA
Jermyn et al., 201754	US	OBS	2014–2016	66	NA	NA	100	NA	NA
Kurek et al., 2017 (COMMIT-HF)61	PL	OBS (matched)	2009–2013	574	NA	84	41	≤35%	71
Landolina et al., 2012 (EVOLVO)62	IT	RCT	2008–2009	200	NA	79	88	≤35%	46
Liberska et al., 201664	PL	OBS	2006–2012	305	62.6	76	NA	≤35%	57
Lindenfeld et al., 2021 (GUIDE-HF)65	US	RCT	2018–2019	1000	NA	63	70	ANY	40
Lüthje et al., 201566	DE	RCT	2007–2011	176	65.9 ± 12.0	77	43	ANY	51
Morgan et al., 2017 (REM-HF)69	UK	RCT	2011–2014	1650	69.5 ± 10.17	86	31	ANY	NA
Mullens et al., 201072	BE/US	OBS	2007–2007	194	62.0 ± 14.0	59	NA	NA	45
Sardu et al., 201685	IT	RCT	2010–2014	191	72.2 ± 7.2	76	55	<35%	NA
Sharif et al., 2022 (SIRONA 2)89	BE	OBS	2019–2021	70	71.0 ± 10.0	71	100	ANY	NA
Shavelle et al., 202090	US	OBS (pre-post)	2014–2017	1200	69 ± 12	62	NA	ANY	41
Smeets et al., 201791	BE	OBS (registry)	2010–2013	282	71 ± 12	82	18	ANY	61
Tajstra et al., 2020 (RESULT)93	PL	RCT	2015–2017	608	NA	81	22	<35%	64
Treskes et al., 202194	BE/NL/CH	NRCT (pre-post)	2018–2019	74	67.2 ± 10.3	84	32	ANY	36
Van Veldhuisen et al., 2011 (DOT-HF)95	EU/AF/ME/AS	RCT	NA	335	64 ± 10	86	37	≤ 35%	56

SD, standard deviation; IQR, interquartile range; NYHA, New York Heart Association classification; LVEF, Left Ventricular Ejection Fraction; US, United States; NL, The Netherlands; DE, Germany; IE, Ireland; IT, Italy; DK, Denmark; AU, Australia; EU, Europe; PL, Poland; UK, United Kingdom; BE, Belgium; AT, Austria; CH, Switzerland; AF, Africa; ME, Middle East; AS, Asia; RCT, randomized controlled trial; OBS, Observational study; NRCT, non-randomized controlled trial; NA, not available.

Only NYHA III patients.

Patient characteristics

The mean age of patients in the non-invasive hTMS studies was 68 ± 13 years; 67.8% were men; 46.6% were classified as NYHA classes III–IV. Of the non-invasive studies, 10 359, 8571, and 4680 patients were included in the TM, STS, and Complex TM categories, respectively (Table 2). In the invasive hTMS studies, the mean age was 66 ± 12 years. Of these patients 75.8% were men, and 47.6% were classified as NYHA classes III–IV. In the invasive hTMS, 9445 and 3494 patients were included in CIED or IHM, respectively (Table 3). Details regarding the prescription of guideline-directed medical therapy (GDMT) for HF are presented in Supplementary data online, Tables S4 and S5.

Clinical efficacy of telemonitoring

All-cause mortality

In 80 studies (both non-invasive and invasive) reporting ACM, 11.1% (2099/18 711) of the patients died in the hTMS group compared with 12.8% (1949/15 231) in the SoC group. Overall, hTMS showed a significant 16% reduction in ACM (OR: 0.84, 95% CI: 0.77–0.93) (Figure 2). Within the treatment effect, the degree of heterogeneity across all studies was considered as not important however, significant (I² = 24%). The funnel plot and Egger’s regression test showed no evidence of publication bias for this endpoint (see Supplementary data online, Figure S1-S6). Non-invasive hTMS showed a 15% reduction in ACM (OR: 0.85, 95% CI: 0.77–0.94, I² = 9%). This effect was primarily driven by the effect of STS. Invasive hTMS showed no significant reduction in mortality (OR: 0.86, 95% CI: 0.70–1.06, I² = 50%) (Figure 2). These results were consistent for both CIED and IHM studies. The results of the sensitivity analyses, in which HRs are obtained, showed similar results compared with the main analyses based on ORs (see Supplementary data online, Figure S7). Results dividing articles based on follow-up times are presented in Supplementary data online, Figure S8.

Figure 2

Forest plot all-cause mortality. TM, telemonitoring; STS, structured telephone support; complex TM, complex telemonitoring; hTMS, home telemonitoring systems; CIED, cardiac implantable electronic devices; IHM, invasive haemodynamic monitoring. *The studies of Mortara et al.⁷¹ and Cleland et al.³⁰ have multiple intervention arms. Therefore, those articles are presented more than once in the forest plot. In the subtotal non-invasive home telemonitoring systems and the total pooled analysis, event rates of each study arm are added together. **From the article of Lindenfeld et al., the post-COVID analysis was used, to avoid bias in observed outcomes due to the COVID pandemic.

First heart failure hospitalization

In 54 studies reporting first HFH, 21.3% (2201/10 352) of the patients receiving hTMS and 25.5% (2263/8883) receiving SoC had at least one HF admission. The pooled non-invasive and invasive studies showed a 19% reduction in first HFHs in patients using hTMS (OR: 0.81, 95% CI 0.74–0.88) (Figure 3). The degree of heterogeneity of these studies was considered to be not important and non-significant (I² = 22%). The funnel plot and Egger’s regression test showed a significant asymmetry (see Supplementary data online, Figure S9–S13). The HF patients using non-invasive hTMS showed a 22% reduction in first HFH compared with SoC, with a moderate degree of heterogeneity (OR: 0.78, 95% CI: 0.70–0.86, I² = 26%). This effect was primarily driven by the STS and TM studies. In contrast, invasive hTMS showed no significant reduction compared with SoC, with a low degree of heterogeneity (OR: 0.89, 95% CI: 0.77–1.03, I² = 5%). These results were consistent within for both CIED and IHM studies. The results of the sensitivity analysis, in which HRs are obtained, showed similar results compared with the main analyses based on ORs (see Supplementary data online, Figure S14). Results dividing articles based on follow-up times are presented in Supplementary data online, Figure S15.

Figure 3

Forest plot first hospitalization. TM, telemonitoring; STS, structured telephone support; complex TM, complex telemonitoring; hTMS, home telemonitoring systems; CIED, cardiac implantable electronic devices; IHM, invasive haemodynamic monitoring. *The studies of Mortara et al.⁷¹ and Cleland et al.³⁰ have multiple intervention arms. Therefore, those articles are presented more than once in the forest plot. In the subtotal non-invasive home telemonitoring systems and the total pooled analysis, event rates of each study arm are added together.

Total heart failure hospitalizations

In 34 studies reporting total HFHs, 3839 HFHs occurred over the course of 10 280 patient-years in patients receiving hTMS compared with 2929 HFHs over the course of 8358 patient-years in the control group. Receiving hTMS was found to be significantly associated with a 15% reduction in the occurrence of HFHs over time (IRR: 0.85, 95% CI 0.76–0.96) (Figure 4). Within the non-invasive studies, the use of hTMS was associated with an 18% reduction in the occurrence of HFH over time (IRR: 0.82, 95% CI 0.70–0.96). In contrast, in invasive studies, no significant effect in the occurrence of HFH was shown (IRR: 0.90, 95% CI 0.74–1.10). Within all invasive studies, the IHM studies showed a significant reduction in the occurrence of HFH, whereas the CIED studies showed no effect. The degree of heterogeneity of both non-invasive and invasive studies was classified as substantial (non-invasive: I² = 70%, invasive: I² = 73%). The funnel plot and Egger’s regression test showed no evidence of publication bias for this outcome (see Supplementary data online, Figure S16–S21).

Figure 4

Forest plot total hospitalizations. TM, telemonitoring; STS, structured telephone support; complex TM, complex telemonitoring; hTMS, home telemonitoring systems; CIED, cardiac implantable electronic devices; IHM, invasive haemodynamic monitoring. *The studies of Mortara et al.⁷¹ have multiple intervention arms. Therefore, those articles are presented more than once in the forest plot. In the subtotal non-invasive home telemonitoring systems and the total pooled analysis, event rates of each study arm are added together. **From the article of Lindenfeld et al., the post-COVID analysis was used, to avoid bias in observed outcomes due to the COVID pandemic.

Risk of bias assessment

Quality assessment was performed using the RoB2 tool and ROBINS-I tool in 73 and 19 studies, respectively. A total of 20.5% of the RCTs were classified as high risk of bias. This was most frequently due to risk of bias in the domain ‘missing outcome data’ and ‘deviations from the intervention’ (see Supplementary data online, Figure S22). A total of 62.5% of the observational articles were classified as serious or critical risk of bias. This was frequently due to the high risk of confounding bias (see Supplementary data online, Figure S23).

Discussion

In this state-of-the-art meta-analysis of 92 studies encompassing 36 549 patients with HF, we show that the use of hTMS modalities in HF patients is associated with a reduction in the risk of mortality, first HFH, and the total HFHs (Structured Graphical Abstract). We found a strong and consistent overall efficacy in reducing all clinical endpoints, with less heterogeneous results than previous meta-analyses on telemonitoring in chronic HF.⁵ Overall, with our findings, the body of evidence for the use of hTMS in the management of these patients is further growing.

Non-invasive home telemonitoring systems

This comprehensive meta-analysis is the first to demonstrate a significant consistent benefit of non-invasive hTMS in HF patients on reducing ACM, first HFH, and the total HFHs. However, considering the separate modalities within non-invasive hTMS, limited power precluded the robustness that is needed to evaluate if each individual modality would reduce total HFHs. When dissecting the results of the different non-invasive hTMS modalities, we demonstrate that TM had a significant reduction in first HFH, while a tendency towards a reduced risk of ACM and total HFHs was observed. This is in contrast to a Cochrane review,⁵ which demonstrated a significant benefit for both ACM and HFH. This difference could be explained by the reclassification of the Tele-HF study from STS to TM.²⁶ The benefits on first HFHs are in line with Inglis et al.⁵ For complex TM, this review was not able to demonstrate a clear benefit, which may be due to the lower number of studies in this category. Nevertheless, complex TM systems may prove beneficial as shown in the TIM-HF2 trial.⁹ Within this RCT, patients were monitored using a combination of TM and STS and provided with 24/7 telemedical support. This complex intervention led to a reduction in the percentage of days lost due to HFH and ACM. Nevertheless, one potential limitation of complex TM systems is that they are labour-intensive and therefore probably not feasible in every healthcare system. The modality described in the TIM-HF2 study requires extra personnel due to the large amount of provided data in combination with continuous accessibility of telemedical support. A desirable solution to this would be automated interpretation of such data, which, obviously, is challenging. In addition, the effects of less labour-intensive alternatives as STS and TM were overall stronger than complex TM. This observation might be explained by differences in the healthcare system and therefore the SoC of the included studies. The CHAMP-HF and CHECK-HF registries, both containing quality-of-care data from two developed western countries (USA and the Netherlands), show substantial differences regarding guideline adherence, prescription levels, and target dose levels of GDMT and devices, which can be related to differences in healthcare system, insurance, and care access.¹⁰⁴

Our results show a significant overall reduction in the incidence of endpoints in patients with HF through the use of non-invasive hTMS. There is, however, some heterogeneity present between studies. On the other hand, the degree of heterogeneity, regarding ACM and first HFH, of the studies included in this meta-analysis is considerably lower as compared with previous meta-analyses. Interestingly, the effect on the outcomes attenuates in the period after publication of Inglis et al.⁵ A potential explanation for this heterogeneity is that studies that include chronic ‘stable’ HF patients (NYHA classes I–II), who experience less events and have a better overall prognosis, will show a smaller effect size on the short term than studies including unstable HF patients who recently had an HF admission and therefore are at a greater risk of a recurrent event. Unfortunately, as these data were not always presented in detail, we were unable to analyse these differences in the context of the current study. Also, we selected many new studies (up to July 2022) especially from the last 5 years with a more structured and integrated approach of hTMS, and this time window is important with the expansion of GDMT between guidelines.

Invasive home telemonitoring systems

This meta-analysis was not able to demonstrate an overall benefit of invasive hTMS on all outcomes. Sensitivity analysis of the different invasive hTMS modalities showed no benefit of CIED monitoring on ACM and HFHs, while IHM showed a significant reduction in total HFHs. The lack of effect of CIED monitoring is important to note. In our meta-analysis, we did not differentiate between CIED with or without impedance measurements to investigate the potential differences in effect. However, a recent meta-analysis by Zito et al.¹⁰⁵ showed no reduction in risk of ACM and HFH using CIED with or without impedance measurements. Additionally, this meta-analysis presented similar results regarding IHM, with several studies showing a remarkably strong result especially those with specifically designed sensors. These findings can be explained by the pathogenesis of HF deterioration. It is well known that increasing filling pressures is one of the first parameters for deterioration of HF, even before overt clinical symptoms are present.¹⁰⁶ By measuring this clinically intuitive parameter (which leads to proactive early interventions), hospitalizations due to HF deterioration can be avoided.¹² These haemodynamic-guided monitoring techniques are very promising. Still, due to their costs, these devices are most likely targeted for those patients who are at higher risk of (re-)admission due to HF and require more intensive monitoring.

Furthermore, there is not only a variation in the identification of distinct pathological parameters between IHM and CIED, but the method of monitoring is often different as well. Pulmonary artery pressures are frequently measured daily to weekly with the use of IHM, and treatment is changed accordingly, while CIED are frequently operated on an alarm basis rather than frequent data monitoring. In addition, alarms are frequently based on less intuitive measurements, such as impendence or algorithms, compared to clinically relevant pressure data.

The IHM was not able to show a benefit on short-term mortality. This is most likely caused by the low power considering the small number of studies and events with a higher level of uncertainty of our data as well as the relative short follow-up time. The type of patients selected is generally sicker or has more advanced HF with NYHA class III and a previous HF hospitalization, with a reduced life expectancy. Several recent drug and device trials could not show a benefit of treatment on mortality in advanced stages of HF and/or against high levels of background therapies. The IHM studies primarily target congestion with modifications in diuretic dosages to prevent decompensation, which potentiates primarily the effects on recurrent HFH. It is unknown whether this translates to indirect benefits on mortality at long term. In the future, the long-term data on IHM will be expanded e.g. with the MONITOR-HF trial.^65,107

Clinical impact and future perspective

This meta-analysis provides support for telemonitoring to be incorporated in HF care. A tailored approach seems necessary in order to lead to a maximal benefit of hTMS with various determinants such as the type of healthcare system, funding, and also characteristics of the patients such as disease severity and symptoms. Patients with more advanced HF (NYHA class III) appear to benefit of a more intensive form of (invasive) monitoring, which could be achieved through IHM with main effects on recurrent HF hospitalizations (targeting congestion), while a patient with a more ‘stable’ HF (NYHA classes I–II) would suffice with the use of non-invasive hTMS, which is simpler and less costly, also considering the enormous patient volumes. This clearly makes sense from a cost-effectiveness perspective, where the most costly method is reserved for the sickest patients who have most to profit from it. In addition, such systems ideally should be adaptable over time, i.e. to intensify when the patient is in a more unstable phase and taper off when the patient is stable. The latter will most likely lead to a higher adherence during prolonged follow-up. Future research should focus on defining these subgroups of patients (based on age, gender, LVEF, NYHA class, stable/unstable aetiology, or other factors) and the effect of the different hTMS modalities on these subgroups. Moreover, the approach will also largely depend on the compatibility with the healthcare system that is already in place. It may also be important to not only focus on detecting HF deterioration, but also to implement a health maintenance strategy.¹⁰⁸

The evidence provided by this meta-analysis supports non-invasive hTMS and invasive hTMS using IHM (but not CIED). Considering IHM, as the CIs of treatment effects are quite wide, more evidence is needed before widespread use of IHM is to be broadly advocated in specific target populations. Still, there is an urge of wider implementation of remote monitoring strategies within clinical practice and healthcare systems. This requires facilities and personnel, which needs to be funded by the healthcare insurances, and also significant advances in IT development and support in hospitals to reduce workload (e.g. with digital technology and artificial intelligence). Many hTMS studies are on top of care, and the field must also work on replacing standard care components by hTMS, self-management at home, and further reduce face-to-face contacts, such as shown by the EVITA-HF study.⁹⁹ To effectuate this, wider implementation needs to be facilitated by the international HF community and guidelines that speak out about their position on hTMS modalities, with the increasing number of studies and data now provided. Also, we need to study and invest in patients, e.g. self-management and involvement in their disease and remote monitoring strategies, which can help in diet, lifestyle, and treatment adherence and close the loop between hospital and patient.

Strengths and limitations

This meta-analysis has several major strengths. The current meta-analysis is the most comprehensive, contemporary, and largest overview of hTMS (with all available modalities) in chronic HF to date including both clinical trial and real-world observational data. To the best of our knowledge, this is the first systematic review and meta-analysis that focuses on both non-invasive and invasive hTMS in contrast to the Cochrane review of 2015,⁵ which only described non-invasive solutions. While we were unable to directly compare non-invasive with invasive hTMS, this study does offer insights into the effectiveness of both modalities. Furthermore, in this meta-analysis, opposed to earlier meta-analyses, we now also differentiated between first HFH and the total HFHs. In our opinion, this manner of analysing the hTMS data is crucial, since both outcomes have different implications and economic impacts.

Several limitations should be mentioned. Firstly, there was still some heterogeneity across studies. Albeit the heterogeneity is decreasing as compared with previous studies, we can specifically observe heterogeneity in the results on the total number of HFHs. The I²-index as a relative measure of heterogeneity, which is not to be considered as an absolute number but as relative categories ranging from <25%, might not be important and >75% considerable heterogeneity. Possible explanations for this degree of heterogeneity are the large variety of hTMS, patient characteristics, and the large variety of HF management between studies. For IHM, the number of studies and events was low, which relates to the I²-index. Attempts were made to minimize this heterogeneity and investigate the effects of the major categories distinguishing between non-invasive and invasive hTMS. The treatment algorithm used in TM strategies could differ across studies, which may have led to the effects demonstrated. As standard care is the comparator, we should acknowledge that the level of standard care varies between studies and in time-period with expanding GDMT. Compared with the Inglis analysis, we observe a decline in heterogeneity of included studies especially in the last 5 years, with many new structured telemonitoring projects.⁵ Secondly, the follow-up times used in the meta-analysis of the total number of HFH were limitedly available across studies. These were calculated, as stated in the methods, which may introduce some bias. Thirdly, distortion of results may be present due to publication bias. However, based on our funnel plots, we assume the risk of publication bias to be low for the majority of the analyses.

Conclusions

Our meta-analysis revealed that overall hTMS are effective in reducing HFH and improve survival. Non-invasive hTMS reduce all endpoints, whereas in invasive hTMS, only IHM reduces recurrent HFHs significantly. Therefore, telemonitoring should be strongly considered and may be integrated in current HF healthcare systems worldwide. For optimal impact, the implementation of hTMS should ultimately be tailored to the individual HF patient and based on compatibility with current healthcare systems.

Acknowledgements

The authors wish to thank Wichor Bramer and Maarten F.M. Engel from the Erasmus University Medical Centre Library for developing and updating the search strategy.

Supplementary data

Supplementary data is available at European Heart Journal online.

Data availability

The data underlying this article can be shared on reasonable request to the corresponding author.

Funding

This work was investigator-initiated and did not receive any external funding. N.S. is supported by a grant from the Dutch Research Council (NWO), grant number: 628.011.214 (STRAP).

Author notes

N.T.B.S., M.T.G., and D.A. contributed equally to this work and are shared first author.

Conflict of interest: D.T. received research grants from Boston Scientific and Biotronik. O.M. received consulting fees from Abbott, AstraZeneca, and Boehringer-Ingelheim. R.d.B. has received research grants and/or fees from AstraZeneca, Abbott, Boehringer-Ingelheim, Cardior Pharmaceuticals GmbH, Ionis Pharmaceuticals, Inc., Novo Nordisk, and Roche; and has had speaker engagements with Abbott, AstraZeneca, Bayer, Bristol Myers Squibb, Novartis, and Roche. R.v.d.B. received an independent research grant and speaker fee from Abbott. J.B. received independent research grant from Abbott for ISS and has had speaker engagement or advisory boards in the past 5 years with Astra Zeneca, Abbott, Boehringer-Ingelheim, Bayer, Daiichi Sankyo, Novartis and Vifor. All other authors declared to have no conflict of interest.