Telemedicine in heart failure—more than nice to have?

So far, the most convincing evidence for a telemonitoring device relates to the implantable CardioMEMS device (Fig. 1; [8]). This device is implanted into the pulmonary artery (PA) and transmits PA pressures to a central service centre. The treating physician receives the results, including the trends over time of these measurements. The physician is advised to react if PA pressure exceeds a certain threshold which suggests congestion, and when it is below the normal range suggesting dehydration. The study was not powered for mortality but showed significant reduction in HF hospitalisation as a result of improved HF management. This effect was maintained in the long term [9]. A comparable rationale was studied in the COMPASS-HF trial. A sensor on a transvenous lead was positioned in the right ventricle (Fig. 2). The primary endpoint rate was reduced by 21%, but this was not statistically significant. There were lower event rates than expected which could make the study underpowered for the primary endpoint [10]. The major limitation of these studies was that the treatment recommendation is very generic, with a plethora of interventions being used (diuretics, vasodilators), at the discretion of the caring physician.

Another form of telemonitoring is part of ICD/CRT devices. Such monitoring has not generated uniform results. The IN-TIME trial reported improved clinical outcomes by using multi-parameter monitoring based on information from an ICD device. By a daily check of several parameters summarised in Tab. 1, a composite clinical score indicating worsening of HF was improved by 37% (odds ratio = 0.63, 95% CI 0.43–0.90) as compared with usual care [11]. The EFFECT study showed a similar effect with a clear reduction of the combined primary endpoint of all-cause mortality and cardiovascular hospitalisation [12]. Encouraging results were also found in the COMMIT-HF trial, a matched cohort study, using different cardiac device brands for the telemonitoring. They observed a long-term effect of significant reduction in mortality (4.9% vs. 22.3%, p < 0.0001), but obviously, this was not a randomised trial [13]. In contrast, The OptiLink HF study could not show any beneficial clinical outcome in advanced HF by using fluid status telemedicine alerts (Fig. 2; [14]). Likewise, the positive effects on mortality and cardiac hospitalisation were not supported in a meta-analysis including 11 RCTs consisting of 5,703 patients; there was only a favourable effect on the number of visits, but no effects on hard clinical outcomes and an increase in unscheduled visits [15]. As to whether the differences can be explained by the use of different devices or different interventions is currently unknown. Therefore, no general recommendation to use monitoring information from implantable devices can at present be made.

In general, individual trials of telemonitoring/telephone support compared with usual care did not consistently report positive results on the primary endpoints (Tab. 1). The large TELE-HF study did not generate any clinical proof for the use of telemonitoring (utilising a telephone-based interactive voice response system collecting daily information on symptoms and weight) [16]. A voice response system was used without direct contact between healthcare providers, possibly resulting in the low adherence of 14% never users and only about half of the patients using the system more than three times per week [16]. Unfortunately, there was no post-hoc analysis to determine if good adherence resulted in better outcomes. Also, the impact of weight changes for monitoring may be overestimated as it was not demonstrated to be effective as a predictive marker of impending decompensation [17]. This is supported by the negative WISH trial that compared a self-measurement of patients’ weight or by an electronic scales with automatic transmission of the results to the clinic. There was a solid mean of 75% (0–100%) of patient compliance, but there was no significant difference in endpoints between the groups or in subgroups [18]. Also the MCCD trial showed no benefit of telemonitoring despite very good adherence of the participants [19]. Again, the system was mainly based on data transmission with very little direct contact with the patients. Further, the TIM-HF group failed to show a positive effect on the primary endpoint of all-cause mortality or composite endpoints comparing usual care with telemonitoring (Fig. 3), but the study was not sufficiently powered [20]. In addition, the CHAT trial showed mixed results with positive effects on the secondary endpoints of all-cause mortality and all-cause hospitalisation with telecommunication, but not on the primary endpoint in HF patients living in rural areas [21]. Very recently, the large TIM-HF2 study found more days alive outside the hospital with the use of structured remote patient management interventions as compared with usual care (Tab. 1; [22]). Taken together, the inhomogeneity of the methods used, the devices applied, the patients included and the intervention performed together with the lack of sufficient statistical power may explain the mixed findings of individual trials regarding the use of telemonitoring. Moreover, there is a lack of direct comparison between different modalities, making it impossible to determine which may be the most effective method.

However, several recent meta-analyses and systematic reviews reported that the use of telemonitoring may improve outcomes [23‐26]. As a consequence of the mixed trials, these meta-analyses included studies with different inclusion criteria. Despite these differences, all meta-analyses reported reduction in mortality and HF-related hospital admissions. In addition, Kruse et al. concluded that telemedicine is effective in customer satisfaction [25] and may increase the sense of security [27]. Also, some but not all studies reported positive effects on quality of life [9, 19, 28]. Still, as many randomised trials were neutral, the recommendation by the ESC HF guidelines is restrictive (i. e. recommendation IIB, level B) [2].

In addition to research design, organisation of healthcare may importantly influence outcomes of healthcare interventions as telemonitoring. By comparing only Dutch studies, we attempt to create a certain level of similarity, as organisation of care is comparable in all parts of the Netherlands. The first randomised study including a significant number of Dutch patients was the TEN-HMS study [29], which compared telemonitoring with more traditional HF care as nursing telephone support and usual care. Telemedicine did not differ from nursing telephone support except for prescription of medication, however both had significantly better results compared with usual care for all endpoints (Tab. 2). The Dutch TEHAF study [30] compared the results of using the Health Buddy®, monitoring signs and symptoms, with usual care. HF hospitalisations and visits to the HF clinic decreased, but the primary endpoint of time to first HF hospitalisation was not significantly improved. The IN-TOUCH study compared an ICT-guided disease management support and an ICT-guided support with additional telemedicine [31]. No significant differences in outcome were found, possibly due to the lack of a usual care group. The e‑Vita study, a prospective three-arm study (usual care; usual care plus the heartfailurematters.org website; these two plus an adjusted care pathway with an interactive platform for disease management (e-Vita platform), replacing routine outpatient consultations with HF nurses), could not show any significant benefit [32]. Lastly, an optimised care program using a telehealth application in a pre-post design [33] during a 12-month follow-up found positive effects on most outcomes. Due to the design and the limited study population, the results should be interpreted with caution.

Taken together, the Dutch studies follow the line of the overall evidence with mixed results, explained by the low power of the studies. Endpoints mostly focus on mortality and care consumption, yet they were not powered to detect differences. Possibly, the high standard of usual care may have influenced the results. The challenge is to detect the important aspects of the systems and how to integrate the systems into the daily care process.

There is limited evidence regarding cost-effectiveness of telemedicine. The reduction of hospitalisation and the increased self-management of patients embodies the potential of cost reduction in healthcare [25]. The incremental cost-effectiveness of the CardioMEMS device is estimated to be $ 13,979 per quality-adjusted life year gained [8]. Klersy et al. describe in their meta-analysis on telemonitoring by cardiac devices a reduction of 44% in hospital visits, without affecting mortality, resulting 15–50% cost saving [15]. In the long term, these interventions were calculated to be cost-neutral [34].

Regarding non-invasive telemonitoring, the effects on costs are even less clear. Blum et al. showed no cost-reduction [19] as there was no positive effect in the study (e. g. readmission/hospitalisation). In contrast, Comín-Colet et al. found a significant reduction in HF and cardiovascular readmission with the use of telemedicine, which resulted in a net decrease in direct hospital costs of € 3,546 per patient per 6 months of follow-up [35]. In the Dutch TEHAF study, the probability of being cost-effective was only 48% (threshold of € 50,000), possibly due to the divergence between participating institutions [36]. Because of the heterogeneity of all the studies, populations and no uniform intervention it is difficult to be conclusive on cost-effectiveness.