Heart failure (HF) represents a multifactorial prevalent syndrome and a frequent cause of hospitalization with significant socioeconomic impact [1, 2, 3]. The volemia status is a key factor in the pathophysiology of this disease, but unfortunately, clinical signs of congestion, such as crackles, jugular vein distention, lower limb edema, or weight gain, may not manifest until substantial volume overload occurs [4]. Timely intervention addressing congestion increases the likelihood of preventing hospital admissions [5, 6]. Given the dynamic nature of HF, monitoring and early identification of congestion are imperative for enhancing prognosis [7, 8, 9].

Multiple biotechnological approaches have been pursued, focusing on volume assessment, such as serum biomarkers (natriuretic peptides or CA-125 antigen) or biophysical parameters [10, 11, 12]. Amidst the latter, bioimpedance (BI)-frequently found in literature as bioimpedance analysis (BIA) or bioimpedance vector analysis (BIVA)-is gaining increasing attention from clinicians and researchers due to its theoretical capability to detect extracellular fluid [13].

BI measures the opposition that living tissues offer to the flow of an alternating current electrical signal. Extracellular fluid expansion, as seen in congestion, generally lead to a decrease in bioimpedance values. Previous studies involving BI assessments have mainly focused on implantable devices or punctual static measurements [14, 15]. Continuous monitoring with BI tools might offer a better and noninvasive evaluation of fluid status and variations in HF [16]. The scope of this systematic review is to delve into the clinical relevance of wearable BI-based monitoring devices for HF.

The database search yielded 5679 records, with 3975 identified as duplicates. Among the remaining 1704 records evaluated by title, 207 progressed to further screening. Only two articles could not be retrieved, leaving 205 records for detailed assessment. Only 10 articles demonstrated sufficient compliance to be included in the review.

The primary reasons for exclusion were as follows (ordered by frequency): non-wearable devices (124), invasive implantable devices (35), letter/opinion papers (13), non-English articles (7), study population not affected by heart failure (6), non-bioimpedance-based wearable devices (4), pre-clinical studies (4), software evaluation studies (4), and study protocols (3). Often, the same study had multiple published articles exploring different aspects or presenting varied results. The article that best filled the inclusion criteria was selected in these cases. While processing the search strategy, nine additional articles were identified through citations and were scrutinized. The selection process is shown in Fig. 1 (Ref. [21]) following the prism flow diagram.

Fig. 1.

PRISMA flow diagram. The article selection process shows the different steps and reviewing processes. The diagram was generated using PRISMA 2020 ShinyApp [21]. BIA, bioimpedance analysis; HF, heart failure; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

The selected ten articles were conducted in North America (United States), Europe (Belgium, Germany, and Spain), and Asia (India and Singapore) from 2012 to 2023. Most studies adopted an observational design characterized by a prospective, non-controlled, and non-randomized approach. Table 1 (Ref. [22, 23, 24, 25, 26, 27, 28, 29, 30, 31]) summarizes the characteristics of the studies included in this systematic review.

The study populations were diverse, comprising a heterogeneous mix of exclusively admitted HF patients (some due to HF and some due to any other cause), initially admitted and later transitioning to ambulatory HF patients, exclusively ambulatory HF patients, and non-HF patients (controls). Analyzed subjects ranged from 2 to 200 per study with a median of 40, gathering 535 individuals. The average age, calculated based on available information, was 66.7 $\pm$ 8.9 years, and males constituted 70.4% of the participants. Left ventricular ejection fraction (LVEF) and New York Heart Association (NYHA) functional classification values were reported for 75.5% and 41.5% of the subjects, with mean $\pm$ SD values of 32.9 $\pm$ 8.1% and 2.5 $\pm$ 0.3, respectively. The mean follow-up duration, available for 89.7% of the analyzed subjects, was 168.3 $\pm$ 130.6 days.

The preferred BI-based wearable devices in eight of the ten studies were vests or Holter-like setups designed to assess transthoracic BI [22, 23, 24, 25, 26, 28, 29, 30]. These devices typically featured four electrodes and measured BI at multiple frequencies. In one of the studies, a two-electrode patch was employed to assess skin BI, while another study utilized a four-electrode anklet to investigate the segmental impedance of the leg, specifically targeting the identification of edema [27, 31].

The outcomes and their definitions displayed notable diversity across the included studies. Many of them, small proof-of-concept prospective studies, aimed to explore the changes in BI in HF-admitted patients undergoing depletive therapy. In 2015, Lee, Squillace, and Smeets [23] found a strong negative correlation in three HF-admitted patients between fluid balance and BI (R² = 0.84 $\pm$ 0.03). In 2020, Reljin et al. [29] collected the transthoracic BI parameters and heart rate variability from the SHIELD study on 44 admitted patients to develop an algorithm that demonstrated an accuracy of 92% in identifying lung congestion. Two years later, Sanchez-Perez and Berkebile [30] observed changes in the resistance at 5 kHz (R_5kHz) and 150 kHz (R_150kHz) ratio, K (R_5kHz:150kHz) assessed through thoracic BI in eight HF patients from admission to discharge, finding a significant increase in K of 0.05 $\pm$ 0.19 (p 0.001).

Four articles focused on event prediction in relation to HF worsening with BI as an early diagnostic tool. Boasting the largest sample size within the review, in 2012, the MUSIC investigators presented an algorithm based on the BI data along with two other physiological parameters (breath index and personalized fluid status) that could predict HF-worsening related events (admission, diuretic up-titration or death) 11.5 $\pm$ 6.0 days ahead with a sensitivity of 63%, a specificity of 93%, and a false positive rate of 0.9 per patient–year [22]. Using a similar approach, in 2016, Cuba-Gyllensten et al. [25] published an enhanced algorithm using only transthoracic-BI that yielded a sensitivity of 60%, a specificity of 96%, and positive and negative predictive values of 10.9% and 99.6%, respectively, outperforming similar algorithms based on weight measurements in terms of predicting HF admissions. One year later, Darling and Dovancescu and colleagues [26], using the BI data from the SENTINEL-HF study and considering the events as diuretic up-titration or HF-related admission, produced with a sensitivity of 87%, specificity of 70%, and an overall accuracy of 72% within the 30-day window before the event. In 2020, the LINK-HF multicenter study that combined skin BI with physical activity and heart and respiratory rates measured using a patch showed a sensitivity of 76% to 88% and a specificity of 85% for detecting re-admissions after HF hospitalization with an alert-to-admission time of 6.5–8.5 days [27].

With this information, without forgetting the diverse nature of the studies and outcomes, we could summarize that the overall sensitivity and specificity described so far in the literature on predicting HF-related events using wearable BI—along with other parameters—measuring devices are around 70.0 (95% CI: 68.8–71.1) and 89.1 (95% CI: 88.3–89.9), respectively. Addressing mortality, apart from the aforementioned MUSIC study, in 2016, Gastelurrutia and Cuba-Gyllensten [24] followed 20 initially admitted HF patients for 18 months after discharge and observed that R₀ (theoretical resistance value at 0 Hz frequency obtained through the Cole-Cole model [32]) on admission was 11.7 $\pm$ 7.12 $\mathrm{\Omega }$ (p = 0.003) lower in patients who died during follow-up. In 2020, Smeets and colleagues [28] studied 36 patients admitted for HF in a coronary care unit, dividing them into two groups depending on whether their transthoracic BI had increased or decreased by discharge. Following discharge, they monitored them for the subsequent 12 months and performed a survival analysis that proved that a decrease in R_80kHz was related to all-cause-mortality with a hazard ratio (HR) of 5.51 (95% CI: 1.55–23.32; p = 0.02) and to the composite outcome all-cause-mortality and HF admission with a HR of 4.96 (95% CI:1.82–14.37; p = 0.01) [28].

In this systematic review, we performed a comprehensive literature research study considering many document types beyond randomized control trials (as we even searched for and found conference papers). The rationale behind this approach was to avoid missing any pertinent insight and to minimize publication bias. However, we understand that certain studies, particularly those associated with industrial development and patent protection, might remain unpublished or inaccessible through our literature search scheme. It is also noticeable that most of the articles included in the review are small-size proof-of-concept studies that present serious concerns while exploring the risk of bias, and no randomized controlled trials were found; hence, conclusions drawn in this review must be cautiously and carefully considered.

Although BI has demonstrated its efficacy in detecting fluid overload and holds substantial evidence in HF management, its integration into clinical practice still faces several challenges [13]. While exploring this parameter, authors commonly find issues with inter- and intra-individual variability and confounding factors such as sex, circadian changes, body composition, position, medications, and skin abnormalities, among others. These factors contribute to the complexity of establishing a consistent reference point for BI measurements, making it challenging to interpret and apply these data in a standardized manner [14].

In a recent meta-analysis, intrathoracic BI data obtained through implantable devices failed to improve HF-related events (HF admission and all-cause mortality) predictions compared with standard care or noninvasive telemonitoring [15]. To interpret these results, the authors propose that pulmonary congestion might manifest as a late-onset feature in HF decompensation, adopting a pathophysiological approach. However, an alternative consideration could be the limited coverage area for impedance exploration provided by these devices. In contrast to implantable devices, wearable devices may offer two primary advantages: noninvasiveness and a broader coverage area for exploration.

Despite the advantages of wearable devices, such as noninvasiveness and continuous monitoring, there are still challenges to be addressed before seamless integration into clinical practice. In the reviewed papers, it is evident that over the years, together with technological development, certain technical issues, such as the quality of measures, signal processing, and the effectiveness of contact electrodes, have improved. However, there remains a need for continued implementation and refinement of this technology to potentially revolutionize the paradigm of HF monitoring and early prediction of decompensations.

Healthcare providers need to consider that the global estimated cost of HF care is USD 108 billion per year, with at least 60% of the cost directly related to admissions. Given these substantial costs, investing in tools that may reduce admissions could be a wise decision for improving prognosis and financially [33].

Wearable devices that measure BI have demonstrated effectiveness in detecting fluid overload and show promise in monitoring HF. However, additional studies are warranted to investigate their potential utility in predicting related events that worsen HF, thereby improving overall prognosis. Further research, including randomized controlled trials in this domain, will contribute to a more comprehensive understanding of the capabilities and clinical implications of these devices in the context of HF management.

The review protocol is publicly available at PROSPERO (CRD42024509914). The articles analyzed during the current study are available through Embase (https://www.embase.com), Cochrane (https://www.cochranelibrary.com), PubMed (https://pubmed.ncbi.nlm.nih.gov), Scopus (https://www.scopus.com) and Web of Science (https://webofscience.com) databases following the steps described in Methods but its access may be restricted depending on researcher’s institution subscriptions.

FJM, LGM, SFS and AY designed the research study. LGM, SFS, PPG, AOF and GH performed the research. LGM and SFS drafted the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Not applicable.

This research was funded by the Instituto de Salud Carlos III through the Real-time monitoring prognostic value of volume with BI test in patients with acute HF (HEART-FAIL VOLUM) project, Grants numbers DTS19/00134 and DTS19/00137, by the Plan Andaluz de Investigación, Desarrollo e Innovación (PAIDI 2020), Consejería de Salud de la Junta de Andalucía, Grant number AT 21_00010_US, and by Caixa Reserch Validate 2022, from La Caixa Foundation (CI22-00287).

The authors declare no conflict of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/j.rcm2509315.