Cost-effectiveness of remote monitoring of implantable cardioverter-defibrillators in France: a meta-analysis and an integrated economic model derived from randomized controlled trials

Abstract

Aims

Cost-effectiveness data on the remote monitoring (RM) of implantable cardioverter-defibrillators (ICDs) compared to the current standard of care (SC) remains limited. This meta-analysis was performed to assess the economic burden, and to develop an integrated economic model evaluating the efficiency of the RM strategy vs. SC in the context of French healthcare.

Methods and results

Randomized controlled trials, comparing RM to SC in patients implanted with ICDs with or without resynchronization therapy (±CRT-D), were identified through a systematic search of scientific literature databases dating from 2005. Seventeen trials (10 229 patients) reporting data on clinical outcomes, quality of life, cost, and/or utility, either as primary or secondary endpoints were identified. Compared to SC, RM resulted in significant reductions in annual costs per patient for direct healthcare costs (seven studies, difference in means −276.1, 95% standard error [SE]: 66.0, I² = 76.3%) and for labour costs (two studies, difference in means −11.3, 95% SE: 1.4, I² = 96.3%). A three-state Markov Model showed that RM resulted in cost-savings of €4142 per patient over a 5-year time horizon, with a quality-adjusted life year (QALY) gain of 0.29. The incremental cost-effectiveness ratio was −14 136 €/QALY, in favour of RM. Furthermore, probabilistic sensitivity analyses confirmed that the RM strategy was dominant over SC in 70% of cases.

Conclusion

Our economic model demonstrates that once implemented, RM of ICD ± CRT-D patients would result in increased effectiveness for lower costs over a 5-year period, compared to the current SC in France.

Graphical Abstract

Remote monitoring, Telemedicine, Implantable cardioverter-defibrillator, Meta-analysis, Cost-effectiveness, Economic model

What’s new?

• This meta-analysis of 17 randomized controlled trials compares the clinical and cost outcomes of implantable cardioverter-defibrillator (ICD) remote monitoring (RM) to the standard of care (SC): quarterly in-clinic follow-ups.

• Cost data obtained is from the French healthcare system.

• Markov modelling showed that the RM of ICDs would result in cost-savings of €4142 per patient over 5-years, with a quality-adjusted life year (QALY) gain of 0.29.

• The incremental cost-effectiveness ratio was −14 136 €/QALY, in favour of RM.

• Once implemented, the RM strategy would be more cost-effective in the long term compared to the current SC.

Introduction

The ageing population and the scarcity of medical resources call for innovative ways of monitoring patients suffering from chronic diseases. Cardiac implantable electronic devices (CIEDs) offer one such opportunity. Over the decades, implantable cardioverter-defibrillators (ICDs) have become a standard therapy in the primary and secondary prevention of sudden cardiac death.¹ The bi-ventricular defibrillator with resynchronization function (CRT-D) has become a treatment of choice for patients with left ventricular dysfunction, with or without heart failure (HF).^1,2 Since their introduction in 1958, the number and complexity of CIEDs have been increasing. Based on Eucomed data, 395 000 pacemakers and 62 000 ICDs were implanted in Europe in 2009.³

Patients with ICDs require high-quality care and regular oversight to ensure the safe and effective performance of their devices. International guidelines recommend that patients be followed-up at 3- to 6-month intervals, depending on the device model and the patient’s clinical condition.^3,4 These in-clinic follow-ups are the current standard of care (SC).

Remote monitoring (RM) systems transmit data at scheduled times, or upon the detection of a particular event (such as an arrhythmia), generating alerts and providing information to clinicians. Their implementation aims to reduce the frequency of routine visits to the clinic, limiting them to annual visits and additional unscheduled visits on a need-to basis, following an alert requiring in-clinic attention.⁵

Although the clinical and logistic benefits of RM have been analysed and lauded over the years, and while CIEDs with RM systems are presently reimbursed by the French healthcare system, the activity of remotely monitoring these devices is not.⁶ Factors such as installation and staff training costs seem to be impeding its implementation. A cost-effectiveness analysis comparing follow-up by RM vs. in-clinic follow-ups is thus warranted to provide objective evidence to enable healthcare authorities to make their decision on the reimbursement of RM activity. Randomized controlled trials (RCTs) evaluating the cost-effectiveness of RM vs. SC have begun to emerge, namely EVOLVO,⁷ EuroEco,⁸ MORE-CARE,⁹ MONITOR-ICD,^10,11 and ECOST.¹²

The objective of this investigation was (i) to perform a systematic review identifying all RCTs comparing RM vs. SC, (ii) to conduct a meta-analysis evaluating clinical outcomes and cost, and (iii) to evaluate the efficiency of the RM strategy, through an integrated economic model within the French context.

Methods

Study selection and data extraction

A systematic search of PubMed, Scopus, ClinicalTrials.gov, and the Cochrane databases was performed to identify RCTs comparing RM to SC in ICD and CRT-D patients. The databases were last accessed on 27 February 2018 and manual searches were conducted between 28 February and 14 March 2018, to include references published between 01 January 2005 and 15 February 2018. Search terms used for database and manual searches are provided in Supplementary material online.

Two authors reviewed titles and abstracts from the searches to screen for articles reporting on clinical, economic or quality of life (QoL) outcomes concerning the RM of ICDs and CRT-Ds. Full-text articles from the first round of abstract screening were further assessed by two authors to select only RCTs that compared RM to SC in a minimum of a 100 ICD and CRT-D patients. RCTs were included if results were published in peer-review journals or as conference abstracts with extractable data. A third reviewer arbitrated disagreements between the two reviewers during the selection processes. Abstracts and articles were excluded for the pre-defined reasons listed in the PRISMA flow diagram in Figure 1.

Figure 1

PRISMA flow diagram of the systematic search and RCT selection. Seventeen RCTs were selected from a total of 1261 references to compare the cost-effectiveness of RM of ICDs against SC, which involves in-clinic follow-ups of ICDs. CRT-D, cardiac resynchronization therapy-defibrillator; ICD, implantable cardioverter-defibrillator; RCT, randomized clinical trial; RM, remote monitoring; SC, standard care.

Data were extracted from the RCTs for the following outcomes: (i) clinical outcomes, including all-cause mortality, cardiovascular (CV) mortality, major cardiovascular adverse events (MCVAE) (including any of the following: any-cause hospitalizations, CV hospitalization, HF hospitalization, device-related hospitalizations, stroke, and surgery), and composite clinical endpoints (two or more of the following: all-cause mortality, CV mortality, MCVAE, CV hospitalization, and any-cause hospitalization); (ii) QoL outcomes, including EQ-5D, SF-36, quality-adjusted life year (QALY), and Minnesota Living with Heart Failure Questionnaire; (iii) costs, including annual costs per patient/year (for hospital visits and device management) and labour costs for RM management. Study quality was evaluated based on the NICE critical appraisal (Supplementary material online, Table S1). Essential details of selected studies and the outcomes extracted from each are presented in Table 1.

Table 1

Selection criteria for patients and data extracted from the selected RCTs

Study name and year	Number of centres, location, and follow-up duration	Selection criteria	Sample size and follow-up routine		Outcomes extracted
			RM	SC
Al-Khatib, 201013	1, USA 12 months	Having an ICD or CRT-D for an approved indication	N = 76 In-clinic visit at 12 months. Patient-initiated transmissions every 3 months	N = 75 In-clinic visits every 3 months	Composite endpoint,a all-cause mortality, MCVAE, EQ-5D
Bohm, 201614	65, Germany 24 months	NYHA Class II or III; LVEF ≤35%; received an ICD or CRT-D within the preceding 3–21 days; meeting one of the three following conditions: HF hospitalization in the last 12 months, IV/oral diuretic treatment within 1 month, increased BNP within 1 month; no chronic renal failure, severe COPD, or heart transplantation planned	N = 505 In-clinic visits every 6 months; RM alerts switched on	N = 497 In-clinic visits every 6 months without RM alerts	Composite endpoint,a mortality, CV mortality MCVAE
Bulava, 201515	1, Czech Republic 37.4 months (mean)	Patients indicated for an ICD, according to ESC guidelines	N = 97 In-clinic visits every 12 months, continuous RM	N = 101 In-clinic visits at 1 and 3 months, then every 6 months	All-cause mortality, MCVAE, Annual costs per patient
Calo, 201316	1, Italy12 months	Approved indication for ICD or CRT-D	N = 117 In-clinic visit at 12 months, remote follow-up at 3, 6, and 9 months	N = 116 In-clinic visits every 3 months	Labour costs
Caravati, 201317	1, Italy 6 months	Patients with a previously implanted ICD for approved indication	N = 75 Not detailed	N = 75 Not detailed	MCVAE, Annual costs per patient
CONNECT, 201118	136, USA 15 months	Implanted with a Medtronic ICD or CRT-D; no permanent AF; no chronic warfarin therapy, not having had a previous ICD, CRT, or PM Stratified by device type	N = 1005 In-clinic visits at 1, 3 and 15 months. Remote transmissions at 6, 9, and 12 months	N = 975 In-clinic visits every 3 months	MCVAE
CONNECT- OptiVol, 201519	1, Germany 15 months	Approved indication for ICD or CRT-D as a new implant or replacement; no permanent AF Stratified by device type	N = 87 In-clinic visits at 1 and 15 months. Remote transmissions at 3, 6, 9, and 12 months	N = 89 In-clinic visits every 3 months	All-cause mortality, MCVAE
DOT-HF, 201120	72: 19 European countries and Hong Kong, India, Israel, Malaysia, Saudi Arabia, Turkey 15 months	NYHA Class II–IV; LVEF ≤35%; HF hospitalization <12 months pre-implantation; enrolment within 6 months of device implantation; OptiVol-enabled device; no scheduled or previous cardiac surgery <90 days pre-enrolment; no MI in the last 40 days; no severe pulmonary or renal disease; no ongoing inotropic drug therapy; no complex congenital heart disease	N = 168 In-clinic visits at 3 and 6 months, then every 6 months until study end or exit. OptiVol alert on	N = 167 In-clinic visits at 3 and 6 months, then every 6 months until study end or exit. OptiVol alert disabled	Composite endpoint,a All-cause mortality, MCVAE
ECOST, 201312	43: France 24 months	Approved indication for an ICD, as a new implant or replacement; NYHA I–III	N = 221 First in-clinic visit 1–3 months after implant, then at months 15 and 27	N = 212 First in clinic visit 1–3 months after implant, then every 6 months	Composite endpoint,a all-cause mortality, CV mortality, MCVAE, annual costs per patient
EuroEco, 20148	17: Belgium, Finland, Germany, Spain, UK, Netherlands 24 months	Approved indication for an ICD with RM abilities, as a new implant or replacement	N = 159 In-clinic visits at 6 weeks post-discharge, and at 12 and 24 months. Continuous RM	N = 144 In-clinic visits 6 weeks post-discharge, and at 12 and 24 months. Interim in-clinic visits as per each centre’s practice	SF-36, labour costs, annual costs per patient
EVOLVO, 20137	6: Italy 16 months	LVEF ≤35%; implanted with a wireless Medtronic ICD or CRT-D with OptiVol algorithm. Stratified by centre and time from implantation (≤6 months and ≥6 months)	N = 99 In-clinic visits at 8 and 16 months; remote transmissions at 4 and 12 months	N = 101 In-clinic visits every 4 months	MCVAE, MLWHF, QALY, annual costs per patient
IN-TIME, 201421	36: Australia, Austria, Czech Republic, Denmark, Germany Israel, Latvia 12 months	HF ≥3 months; indication for dual-chamber ICD or CRT-D; NYHA II or III for 1 month prior to screening; LVEF ≤35% within 3 months prior to screening; no uncontrolled hypertension, permanent AF, cardiomyopathies or acute myocarditis. stratified by centre	N = 333 In-clinic visit at 12 months; RM every 6 months	N = 331 In-clinic visits every 6 months	All-cause mortality, CV mortality, MCVAE
MONITOR-ICD, 201710,11	13, Germany 24 months	Approved indication for an ICD, as a new implant or replacement	N = 199 In-clinic visits at 1, 12, and 24 months. Remote follow-up at 3, 6, and 18 months. QoL at implantation, 12 and 24 months	N = 203 In clinic visits at 1, 3, 6 months, then every 6 months. QoL at implantation, 12 and 24 months	All-cause mortality, annual costs per patient
MORE-CARE I, 201322	32: France, Hungary, Israel, Italy, Spain, Switzerland 12 months	Approved indication for CRT-D (LVEF ≤35%, NYHA III–IV, QRS ≥120 ms); implanted within the last 8 weeks; device providing fluid accumulation diagnostics; no permanent AT or AF; no prior CRT or CRT-D implantation	N = 76 In-clinic visits at 1 and 8 months, Remote follow-up at 4 and 12 months	N = 72 In-clinic visit at 1 month, then every 4 months	All-cause mortality, CV mortality, MLWHF
MORE-CARE II, 20169	61: 9 European countries and Israel Interim results available only		N = 437 Same as for MORE-CARE I	N = 428 Same as for MORE-CARE I	Composite endpoint,a all-cause mortality, CV mortality, MCVAE, MLWHF, annual costs per patient
REM-HF, 201723	9, UK Minimum 24 months	NYHA Class II–IV; ICD, CRT-D, or CRT-P implanted for ≥6 months; no device implantation or change <30 days pre-enrolment; no acute MI or cardiac surgery <3 months pre-enrolment	N = 826 Contact with patients at 3, 6, 12, and 24 months. Continuous RM	N = 824 Contact with patients at 3, 6, 12, and 24 months. Interim in-clinic visits as per each centre’s practice	Composite endpoint,a all-cause mortality, CV mortality, MCVAE
TRUST, 201024	102: USA 12 months	Implanted within the last 45 days or being considered for implantation with a BIOTRONIK ICD for a Class I or II indication; not PM dependent	N = 908 In-clinic visits at 3 and 15 months; remote follow-up at 6, 9, and 12 months	N = 431 In-clinic visits every 3 months	All-cause mortality, CV mortality, MCVAE

Patients in all studies were 18 years or older and gave their informed consent to participate.

AF, atrial fibrillation; AT, atrial tachyarrhythmia; CRT, cardiac resynchronization therapy; CRT-D, cardiac resynchronization therapy-defibrillator; CRT-P, cardiac resynchronization therapy-pacemaker; CV, cardiovascular; ESC, European Society of Cardiology; ICD, implantable cardioverter-defibrillator; LVEF, left ventricular ejection fraction; MCVAE, major cardiovascular adverse event; MI, myocardial infarction; MLWHF, Minnesota Living With Heart Failure questionnaire; NYHA, New York Heart Association; PM, pacemaker; QALY, quality-adjusted life year; QoL, quality of life; RCT, randomized clinical trial; RM, remote monitoring; SC, standard care.

a

Two or more of any of the following: all-cause mortality, MCVAE, CV hospitalization, and/or device-related hospitalization.

Meta-analysis and statistics

The Comprehensive Meta-Analysis software (Version 2.2) was used for this meta-analysis.

In the meta-analyses of dichotomous outcomes (i.e. all-cause mortality or MCVAE), odds ratios were calculated with 95% confidence intervals (CIs) using 2 × 2 tables from the original articles, to evaluate the safety of RM between the groups, whenever possible. Pooled-effect sizes with 95% CIs were calculated using a random effects model, following the method of Der Simonian and Laird, which is based on the inverse-variance approach. In the meta-analyses of continuous outcomes (i.e. costs), mean values and standard deviations (SD), or changes from baseline and SD were used. Where SDs were unavailable, within-study imputations to handle missing information from published studies were performed. The available means and SDs from all the studies with complete information were used to calculate the coefficient of variation.²⁵ The difference in means with 95% CIs were calculated using the inverse-variance random-effects method.

Heterogeneity was determined using the I² test, which measures the percentage of total variation across studies. I² (%) = 100 × (Q-df)/Q, where Q is Cochrane’s heterogeneity statistic and df signifies the degree of freedom. Negative values for I² were set to zero, and an I² ≥ 50% was considered to be of substantial heterogeneity. Significance was set at P = 0.05.

Cost-effectiveness model

A cost-utility analysis was performed from the perspective of the healthcare system in France and over a 5-year time horizon. The economic evaluation was carried out according to the French Health Authority guidelines.²⁶

A Markov multi-state model (TreeAge Pro 2016) with 1-month cycle was employed, in which each patient existed in one of three mutually exclusive states: (i) stable outpatient (the initial state), (ii) CV hospitalization, or (iii) dead (absorbing state). For each month, each state was associated with a utility value, resource use, and cost components. All patients entered the model at the moment of hospital discharge, after having received the ICD ± CRT-D. Patients accrued hospitalizations, costs and utilities in each 1-month cycle until they died or completed 60 cycles.

The same model was applied to RM and SC patients, with only differing inputs for transition probabilities, costs¹² and utility^7,27 (Table 2). All cost data were extracted from the French Information Systems Medicalisation Program (PMSI) and the French RCT, ECOST,¹² which compared the cost-effectiveness of RM vs. ambulatory follow-ups of ICDs. The costs for each group in the ECOST trial were obtained from the original billing documents issued by the French health insurance system, including (i) direct costs of ambulatory ICD follow-ups and associated transportation expenses, (ii) direct costs of ICD-unrelated ambulatory follow-ups, CV treatments and procedures, and (iii) direct hospital costs for management of CV disorders. The cost of ICD per patient was calculated as a function of the remaining device longevity at the end of the study (Table 2), which was evaluated by the slope of battery depletion over time, using a linear regression model. In case of death, the full price of a partially unused device was entered in the model. Equipment costs for telemonitoring that were reimbursed by private health insurance were also considered. However, professional or physician fees for RM consultations were not included.

Table 2

Summary of base case inputs over 5-year time horizon for Markov multi-state model

Input	RM	SC	Data source
Transition probabilities
Probability of death	0.004	0.004125	ECOST trial12
Probability of CV hospitalization	0.0116	0.0129	ECOST trial12
Probability of in-hospital death (during CV hospitalization)	0.029	0.029	PMSI 2017 (mortality associated with CMD 05)
Costs (€)
Per stable outpatient per month	141 ± 94	163 ± 85	ECOST trial12
CV hospitalization per stay	2829 ± 6382	3549 ± 9714	ECOST trial12
Cost of RM device	864 (lifetime)	–	LPP 2018
ICD cost after 60 months of use	12 960 ± 4549	14 144 ± 6944	Based on the remaining longevity and calculated using a linear regression
Cost of ICD (based on the remaining longevity of the devices after 27 months)	5832 ± 2047	6365 ± 3125	ECOST trial12
Cost of ICD device	12 000 single-chamber		PMSI 2018
	14 000 double-chamber
Utilities
1–16 months after ICD implantation	0.793 ± 0.179	0.737 ± 0.234	EVOLVO7
>Month 16	0.754 ± 0.275	0.711 ± 0.305	EVOLVO7
Utility decrease after each CV hospitalization	0.024	0.024	Assumption, based on Göhler et al.27

CV, cardiovascular; ICD, implantable cardioverter-defibrillator; PMSI, French Information System Medicalisation Program; RM, remote monitoring; SC, standard care.

QALY is the effectiveness metric used in this analysis. Each health state is associated with a utility value from 0 to 1, where 0 represents ‘death’ and 1 represents ‘ideal health’. QALYs were computed by aggregating the total time spent in each health state and applying the appropriate utility weight. The model outputs include mean QALY, mean costs and the incremental cost-effectiveness ratio (ICER) expressed in €/QALY. To estimate the present value of future costs and benefits, we adhered to the French recommendations for cost-effectiveness studies by using an annual discount rate of 4% (0.33% monthly) for both, costs and efficacy parameters.²⁶

Base-case probabilistic sensitivity analysis (PSA) was undertaken. The multistate probabilistic model was used to extrapolate survival, utility and resources over the total lifetime of 1000 hypothetical patients.

Results

Search results

Twenty-three publications referencing 17 RCTs were identified from a total of 1261 non-duplicate citations (the selection process is detailed in Figure 1). The quality of most studies was assessed to be good as per the NICE appraisal guideline. Exceptions were Calo et al.,¹⁶ Caravati et al.,¹⁷ and Bulava et al.¹⁵ which lacked clear descriptions of the study design, objectives, and participant flow, EuroEco⁸ in which the randomization strategy was not clearly reported and baseline characteristics were imbalanced between SC and RM groups, and the DOT-HF²⁰ and MORE CARE II⁹ studies, which reported interim results and outcomes from the per-protocol population, respectively.

Study population

Patient baseline characteristics and links to the public trial registry (where possible) for each selected RCT are presented in Supplementary material online, Table S2; 10 229 patients were included in the 17 RCTs, which were published between 2010 and 2017. Population sizes ranged from 150 to 1980, and 12 out of 17 studies were multicentric. The mean follow-up duration ranged from 6 to 37.4 months. Three studies (EuroEco,⁸ MORE-CARE II,⁹ and DOT-HF²⁰) were terminated early due to low-enrolment rates and presented interim data. In brief, the mean age of patients in the trials ranged from 54.7 to 69.5 years. The proportion of males ranged from 72.0% to 88.2%. Mean left ventricular ejection fraction (LVEF) at inclusion varied from 20.4% to 40.0% and the distribution of New York Heart Association (NYHA) classes was also noted (Supplementary material online, Table S2).

Clinical outcomes and annual costs per patient

Random and fixed effects models were used to study the impact of RM vs. SC on clinical outcomes for binary data, while continuous numerical scores were compared using differences in mean.

All-cause mortality was computed for 12 studies, where no significant difference was observed between RM and SC groups (P = 0.053), for both, the random and fixed effects models. Heterogeneity amongst the studies was low (I² = 32.4%, P = 0.131). Similar results were observed for CV mortality (P = 0.201), which was computed for seven studies (I² = 27.3%, P = 0.220). Similarly, no differences were found for MCVAE (computed for 13 studies, P = 0.302) and for composite CV endpoints (computed for six studies, P = 0.885). Heterogeneity in these models was negligible (3.41%, P = 0.395 and 0.0%, P = 0.760, respectively). The Forest plots for the meta-analysis of these clinical outcomes are presented in Supplementary material online, Figure S1.

A significant difference was found in the annual costs per patient (€), computed for seven studies, all performed in Europe (Figure 2A). Remote monitoring was associated with a significant decrease in annual costs per patient of 276 € (SE = 66 €), in both mixed and random effect models, but with high heterogeneity among studies (I² = 76.3%). Evaluated costs included hospitalization, visits, and device management costs. A sensitivity analysis was performed to address this heterogeneity by excluding lower quality studies from the meta-analysis (four studies with <2 points, according to the NICE appraisal—Supplementary material online, Table S2): RM was still found to be associated with higher per patient cost-savings (Supplementary material online, Figure S2), albeit the difference was not significant (−852 € per patient, SE 482, I² = 33.6%, P = 0.22). Labour costs were also computed from two studies (Figure 2B). An annual savings of about 11 € (SE = 1.41 €) per patient favouring RM was observed (I² = 96.3%), which was found to be statistically significant using the fixed-effect model, but not with the random-effect model.

Figure 2

Forest plot representing meta-analyses of costs from the selected RCTs. (A) Annual cost per patient was derived from seven studies enrolling 2551 patients. The elements of annual cost evaluation differed in each study but consisted of costs relating to hospitalization, transportation, in-clinic, and/or remote follow-up visits, unscheduled visits, outpatient care, therapies, and device management. (B) Labour costs per patient/year were obtained from only two studies, comprising 536 patients and included costs incurred by hospital personnel during in-clinic and remote follow-ups. CI, confidence interval; RCT, randomized clinical trial; RM, remote monitoring; SC, standard care.

Cost-effectiveness analysis

Supplementary material online, Figure S3 shows the Markov three-state model structure, over a 5-year time horizon and Table 2 summarizes the input data (transition probabilities, costs, and utilities). In the base case, the model showed that RM provided a cost-saving of 4142.32 € and a QALY gain of 0.29 compared to SC over 5 years, thus demonstrating that RM is the dominant strategy (Table 3, Figure 3). The ICER was −14 135.80 €/QALY. Most costs were attributable to device implant and CV hospitalizations. In addition, PSA confirmed that RM was dominant over SC in 70% of cases (Figure 4).

Figure 3

Cost-effectiveness analysis showing that the RM of ICDs is a dominant strategy compared to SC.

ICD, implantable cardioverter-defibrillator; QALY, quality-adjusted life year; RM, remote monitoring; SC, standard care.

Figure 4

Incremental costs vs. incremental effectiveness (QALYs) from 10 000 PSA simulations. Results reflect the effect of uncertainty in the model on differences in costs and QALYs for patients undergoing RM relative to those having SC. The PSA shows that RM is cost-effective in 70% of cases. PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year; RM, remote monitoring; SC, standard care.

Table 3

Summary of base case results and sensitivity analyses over 5-year time horizon

Base case deterministic analysis over 5 years	Mean costs (€)	Mean QALY
Remote monitoring	103 296.0	3.06
Standard of care	107 438.4	2.76
Difference	4142.32	−0.29
ICER (€/QALY)	−14 135.8 €/QALY
Probabilistic sensitivity analysis—Monte Carlo simulation
Remote monitoring	100 648.7 ± 5892.5	3.01 ± 0.10
Standard of care	105 734.8 ± 7153.0	2.90 ± 0.10
ICER (€/QALY)	−48 196.8 ± 86 886 €/QALY

ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year.

Discussion

The objectives of this research were (i) to assess the efficacy and cost-effectiveness of RM compared to SC through a systematic review and robust meta-analysis and (ii) to evaluate its long-term efficiency from a healthcare perspective in the French context.

The meta-analysis found no difference in clinical outcomes between the RM and SC strategies. These results are consistent with a previous meta-analysis by Parthiban et al.,²⁸ who concluded that RM is non-inferior to SC in terms of survival and safety, but with significant improvements in inappropriate shock reduction and early event detection. A meta-analysis of eight studies in a French Health Technology Assessment (HTA),²⁹ demonstrated that RM use resulted in reduced data transmission times, inappropriate shocks and total in-clinic consultations, without increasing the risk of serious adverse events. Indeed, early event detection would allow personalised management in order to prevent clinical deterioration (such as early anticoagulation in patients with atrial fibrillation), or for device programming optimization to prevent inappropriate shocks (demonstrated in Bulava 2015¹⁵ and ECOST¹²).

Our meta-analyses demonstrated a significant difference in costs per patient, favouring the RM approach. Remote monitoring may allow for potentially continuous patient monitoring with a reduced number of consultations with the specialist. This has been observed in Bulava 2015,¹⁵ EVOLVO,⁷ IN-TIME,²¹ and MORE-CARE.^9,22 Other major factors that could drive cost-savings from a healthcare perspective include the reduced occurrence of inappropriate shocks, which in turn could slow down ICD battery depletion (Bulava 2015¹⁵ and ECOST¹²) and reduced labour costs (EuroEco⁸ and Calo et al.¹⁶). Nevertheless, effect-size heterogeneity for annual and labour costs per patient are very high, implying that net cost-savings from RM could vary in magnitude across diverse healthcare systems. Sensitivity analyses for annual costs per patient further demonstrates the extent of this heterogeneity. Trials that were retained in the sensitivity meta-analysis (Supplementary material online, Figure S2) were multicentre studies performed in a single country (ECOST in France, EVOLVO in Italy and MONITOR-ICD in Germany). We propose that differences in how costs were calculated in each study, inter-nation variability in direct hospital costs, variations in standard clinical practices and differences in study sizes are some factors at the origin of this heterogeneity. While the RM strategy is indeed shown to be more cost-saving than SC, the generalisability of these results ought to be interpreted with caution from one system to another.

Cost-effectiveness was modelled with data mainly from French sources, revealing that RM could provide greater effectiveness for lower costs than SC in France. Our cost-utility model generated a negative ICER of −14 135.80 €/QALY: ICD patients managed through RM gained 0.29 QALY more than those under SC over a 5-year time horizon, with savings of 4142 € per patient. These results are consistent with a meta-analysis in which cost-savings from RM compared to SC in HF management ranged from 300 to 1000 €, with a QALY gain of 0.06.³⁰ Our findings from the model are robust: PSA confirmed RM dominance over SC in 70% of cases (Figure 4). In addition to the hospital costs related to CV events, we considered direct non-hospital costs (i.e. those related to device management, ICD-related ambulatory visits, transportation for ambulatory visits, CV procedures, ICD-unrelated ambulatory visits, and CV treatments). Nevertheless, utility values were not reported in the ECOST study or in any other French study conducted on ICD patients managed with or without RM (Table 2). These were, therefore, derived from the Italian EVOLVO study,⁷ assuming that population and clinical practices are similar between these two countries. Mean baseline utility was slightly imbalanced between RM and SC in EVOLVO, but not significantly different (0.793 vs. 0.737; P = 0.08, Table 2). There was also a statistically non-significant difference at 16 months after ICD implantation (P = 0.32).

Results from cost-utility analyses have clear implications to inform policy makers and payers. A European survey revealed that physicians consider RM to be a clinically useful technology, which has led to significant benefits for patients and a reduction in in-office consultations.³¹ Indeed, the implementation of patient telemonitoring is of growing interest for many European healthcare systems. In France, a decree published in October 2018 defines the objectives, scope and criteria for participating in a nationwide pilot assessment of RM for ICD and pacemaker users over a period of 4 years. The objective of this investigation is to fix provisional tariffs for acts of RM for healthcare professionals, thus enabling a medico-economic evaluation to be presented to Parliament in June 2021.

In Germany, cost-benefit assessments of RM vs. SC in patients with advanced HF have been ongoing since 2016, and in September 2019 a final HTA was published.³² A final decision on RM reimbursement is expected to be delivered in early 2021.³³ In Italy, healthcare decisions are entrusted to individual regional entities. Veneto is the first and only Italian region to have evaluated the economic benefits of ICD RM (HTA of 18 November 2013³⁴) and to approve a regional decree recognising RM as the preferable procedure (DDR 95—7 August 2019), while issuing guidelines on the type of ICDs to be used, based on patient characteristics.

However, the situation is different in other countries. In the UK, for example, there is currently no clear guidance on the use and reimbursement of RM at the national level, and reimbursement policies vary across healthcare centres.³⁵ Contrary to the above countries, a cost-consequence analysis of continuous RM of CIEDs in the UK found that the RM strategy was cost neutral from the UK Health Service perspective.³⁶

It would have been interesting to have more data on workload, time-spent by labour and direct costs associated with setting up an RM system. Especially because high implementation costs are suspected to be one of the factors preventing the rapid acceptance of RM in routine practice across countries. Non-reimbursement itself also discourages the adoption of RM, creating a cycle of doubt. In light of the results of our study, this cycle should be broken. The data presented herein provide objective evidence that once set-up, RM is superior to SC, and that its reimbursement would be beneficial to the already burdened healthcare system.

Limitations

There are other limitations in this model and meta-analysis that deserve consideration, in addition to the limitations discussed earlier. First, due to limited QALY and EQ-5D data available from trials, meta-analysis on QoL measures was not possible. Second, data on indirect costs were lacking in the French study and other European studies, therefore, it was not possible to consider productivity or income losses for patients. Nevertheless, the mean age of the studied patients was >60 years, which is roughly equivalent to retirement ages, so indirect costs might have a negligible impact on the model. Finally, although medical staff used additional time to manage event alerts in the RM group, the ECOST study did not include the pricing for remote follow-up, as this activity is not reimbursed in France. However, ambulatory follow-ups can be highly time-consuming, with a mean duration of 27 min, as estimated from the results of a European survey,³⁷ while RM has been shown to be time saving.^16,38,39 An analysis of the REFORM trial found that RM saved 81 h of medical time per year per 100 patients followed;⁴⁰ this time is more than compensated by the decreased number of ambulatory visits. Taken together, we assumed that the additional cost of performing event alert management would minimally impact our results.

Conclusions

The findings from our selected RCTs provide strong evidence supporting the safety and effectiveness of RM, while suggesting potential benefit with technologies using regular transmission verification. The results from our cost-effectiveness model demonstrate that this technology is cost-effective and a dominant solution over in-clinic management, particularly in France but also in Europe. These findings have clear implications to inform policy makers and payers.

Future investigations should focus on the set-up costs of RM through large-scale economic evaluations tailored to patients with ICDs, and to maximise the clinical and economic benefit of this technology.

Supplementary material

Supplementary material is available at Europace online.

Acknowledgements

The authors thank Dr Alexandra Kumichel and Dr Gabriella Passoni of ClinSearch for their assistance in preparing this manuscript.

Funding

No external funding was received for this research.

Conflict of interest: ClinSearch is a Contract Research Organization (CRO) that has (or had) collaborations with Medtronic, ZOLL Medical, Johnson and Johnson, and Chronolife, France.