Hospitalisations for congestion are a hallmark of heart failure (HF) with each admission being of prognostic importance (1, 2). Implantable haemodynamic monitor (IHM) studies have demonstrated that subclinical alterations in physiology occur weeks before the development of overt clinical signs and symptoms that lead a patient to present to their care provider including a rise in intra-cardiopulmonary pressures and pulmonary congestion (3–5). Body weight is often monitored as an outpatient for the development of congestion but may not accurately predict decompensation (6, 7). IHMs such as pulmonary-artery sensors may reduce hospitalisations for HF but have yet to demonstrate a mortality benefit and are costly, requiring a dedicated invasive procedure (8). IHM-guided care received a modest class IIb recommendation in the 2021 European Society of Cardiology guidelines for HF (9). Wearable devices (wearables) offer a potential non-invasive alternative method of detecting pulmonary congestion and changes in physiological parameters (10, 11). Wearables could be applied readily by patients or carers with no requirement for an invasive implant and used by patients with HF across the range of ejection fraction (EF). In this meta-analysis, we examined the efficacy of these novel technologies that measure pulmonary congestion to reduce hospitalisations for HF, worsening HF events [hospitalisation for HF or emergency department (ED) visits for HF therapy] and mortality, compared with standard, unmonitored care.

This is the first meta-analysis to pool hospitalisation or mortality events from trials examining the effectiveness of wearable devices in HF. The main results support the use of wearable devices to guide care in patients with symptomatic HF irrespective of EF. Wearable device guided treatment provided a 41% reduction in the risk of hospitalisation for HF and a 26% reduction in all-cause mortality compared with standard care alone. Such substantial reductions in clinically important endpoints compare favourably alongside the benefits observed in randomised controlled trials of contemporary standards of HF care, including dapagliflozin (30% reduction in worsening HF events and 17% reduction in mortality) (29) and sacubitril-valsartan (21% reduction in hospitalisation for HF and 16% in mortality) (30). The findings of this meta-analysis also exceed the 28% reduction in hospitalisation for HF observed in the CardioMEMS Heart Sensor Allows Monitoring of Pressure to Improve Outcomes in NYHA Class III HF Patients (CHAMPION) (8) IHM trial. In the recent haemodynamic-GUIDEed management of Heart Failure (GUIDE-HF) randomised trial, neither hospitalisations for HF or mortality were reduced by IHM-guided care (31). Placed alongside CHAMPION and GUIDE-HF, there are some important differences between the IHM trials and the wearable studies included in this analysis. Firstly, CHAMPION and GUIDE-HF were multi-centre, randomised studies enrolling people with ambulatory HF (NYHA class III in CHAMPION, class II–IV in GUIDE-HF) and while prior hospitalisation for HF was an inclusion criterion in these trials, neither purposefully enrolled patients during an acute event. A formal comparison between an implanted and wearable device in a randomised trial is unlikely to ever happen. However, the lower risk option of a non-invasive wearable device that is proven to reduce clinically meaningful events would offer a sensible monitoring strategy to most clinicians.

The magnitude of relative benefit afforded to patients who were managed with wearable-based care compared with standard care alone is particularly notable given the high-risk profile of these people. In all four of the wearable studies included in this analysis, people were recruited either during a hospitalisation or within 10 days of discharge, a prognostically important period. In the three studies that reported NT-proBNP, arguably the most powerful risk predicting variable in HF, natriuretic peptide levels were markedly elevated. The rate of (re)hospitalisation in the control arms of the ReDS and IMPEDANCE-HF trials was between 50 and 94 per 100 person-years respectively, exceeding those rates observed in the same groups in the CHAMPION (68 per 100 person-years) and GUIDE-HF (49.7 per 100 person-years) trials, and considerably higher than in other contemporary trials such as the Dapagliflozin and Prevention of Adverse Outcomes in Heart Failure (DAPA-HF) trial, where the rate of hospitalisations for HF was 5–10 times lower (9.8 events per 100-person years) (29). The elevated event rate in this meta-analysis underscores how highly selected these patients were but also how effective a strategy of wearable-guided care may be to protect such a group at their most clinically vulnerable, a potentially valuable tool when transitioning care from the acute to community setting.

Several factors need to be considered to determine whether the wearable devices included in this analysis are practical monitoring options. Firstly, the duration a device was worn must be sufficient to capture meaningful amounts of data, balanced against patient comfort to ensure compliance using it. Two devices (the ReDS and LI systems) were worn in a healthcare setting for minutes at a time, minimising patient burden but healthcare providers were involved and the practicality of home use by patients was not explored. The Zoll HFMS device was used as intended within the BMAD study, for twenty-four hours a day and at-home. Data transmission rates were not reported for the BMAD trial but in wearable and IHM studies to date, adherence to data transmissions were approximately 90% of study days (8, 11). An indicator of whether a wearable device that is applied to a person's chest is likely to be adopted by patients is whether both men and women consented for the investigational study. The generalisability of wearable devices other than those included in this analysis has been reduced by near exclusive male enrolment in other studies (11). In this meta-analysis, a pooled average of 35% of participants were female, lower than real-world cohorts of people admitted to hospital with HF (32) but higher than the proportions of females recruited to other HF trials in which approximately a quarter were women (29, 30).

The included trials enrolled different populations of people with HF (patients in hospital, patients who were recently discharged, and routinely monitored ambulatory patients with HF). As the wearable devices were examined in patients in different settings and clinical status the generalisability of the meta-analysis findings are broadened. The IMPEDANCE-HF trial recruited high-risk patients with an EF ≤35%, whereas the BMAD, ReDS and ReDS SAFE trials enrolled patients with HF across the range of EF. A combined 46% of participants in the BMAD and ReDS studies had an EF >40%. The majority of patients with an EF >40% do not have an indication for an implantable device such as a defibrillator and so would not receive device-related HF diagnostics such as the COMPASS algorithm (Medtronic, Minneasota, USA). While sub-analyses of the CHAMPION study indicated that people with a preserved EF had a reduction in hospitalisation for HF, these data were limited by small numbers of patients and events (33). Among people with an EF > 40%, IHM-guided care did not reduce hospitalisations in the primary analysis in GUIDE-HF and a recent meta-analysis reported uncertainty regarding the benefits of IHM-guided care in people with an EF ≥50%1 (31, 34). Until wearable device studies report outcomes according to EF classification, it is difficult to conclude whether wearables that detect pulmonary congestion are an effective tool in caring for different HF populations. As people with preserved or milder impairment in left ventricular systolic function were participants in three of the four included studies, and given the ready application of these devices without an invasive procedure, future wearable trials should continue to actively recruit participants across the range of EF and subsequently report outcomes according to EF. Patients with both advanced and mild symptoms at baseline were recruited across the different trials. In the ReDS-SAFE HF trial, 99% of patients were NYHA class III/IV, compared with 52.5% in IMPEDANCE-HF. Rates of guideline-directed medical therapy for HF also differed across the trials, notably the use of sodium glucose co-transporter 2 (SGLT2) inhibitors and mineralocorticoid receptor antagonists (MRAs). Only the patients in BMAT study, which was conducted after the DAPA-HF (29) and Emperor-Reduced trials (35) were reported, were taking SGLT2 inhibitors at baseline. Rates of MRA use also differed substantially between studies [9.6% in ReDS (17) and 61.5% in IMPEDENCE-HF (18)]. Without individual patient-level data, which were not available for analysis, we were however unable to test for any interactions between major patient subgroups (e.g., symptoms severity, treatments, EF) and the effect of treatment guided by a wearable device or not.

This meta-analysis has several limitations. Notably, the four included studies assessed three different wearable devices for detecting pulmonary congestion, limiting direct comparability across studies. Additionally, only one study (BMAD) evaluated the device's effectiveness in a home setting, where patients used it for remote monitoring. Therefore, the feasibility of home use demonstrated in BMAD cannot be reliably extrapolated to the other devices. While both the ReDS and IMPEDANCE-HF studies were randomised controlled trials, both were conducted at single centres with expertise in the management of HF. The BMAD study was a non-randomised concurrent-control trial. Nonetheless, it is well established that patients hospitalized with HF are frequently discharged with residual congestion—a factor strongly associated with increased mortality and rehospitalization (36). Wearable technologies that detect pulmonary congestion in this vulnerable window offer an opportunity for timely intervention. Artificial intelligence (AI)-enabled devices offer a paradigm-shift in the capabilities of remote HF management. The HEARTFELT device, examined in the FOOT study, used AI to analyse 3-dimensional images of pedal oedema, presenting a novel non-invasive method to monitor for decompensation (36). The integration of AI into wearable devices and remote monitoring technologies in general is likely to develop rapidly, and as with any emerging technology should be examined in randomised clinical trials to confirm their role in the management of people with HF.

HF treatment guided by wearable devices that detect pulmonary congestion reduced hospitalisations for HF and mortality in patients with HF across the range of EF. This strategy was effective in patients who had features of adverse prognosis, including a recent hospitalisation, and indicate that wearable monitoring may provide greater protection against adverse events when discharging care from the acute to home setting.