Scope and signifiance

Cardiovascular disease accounts for over 18 million deaths annually, yet its monitoring remains anchored in episodic, clinic-based assessments a model fundamentally at odds with the continuous, dynamic biology of the conditions it tracks. Consumer-grade wearable technology, now embedded in the daily lives of hundreds of millions, offers a credible and scalable alternative. Hughes et al. [1], in a State-of-the-Art Review in the European Heart Journal, provide the most comprehensive synthesis to date of wearable devices in cardiovascular medicine. Their review is unusual in combining rigorous clinical evidence appraisal spanning the full prevention continuum with practical guidance for clinicians navigating wearable data in real-world practice. This commentary reads the review closely and contextualises it against four additional publications that sharpen its principal arguments: the TRUE-HF study [2] and companion editorial [3] providing the first prospective evidence linking smartwatch signals to hard HF outcomes; the review by Lowres et al. [4] on wearable-based AF detection from the same journal; and the feasibility study by Choi et al. [5] on cuffless blood pressure monitoring via PPG on a Samsung Galaxy Watch.

Sensor technology and regulatory framework

A distinctive strength of Hughes et al. [1] is its thorough grounding in the technological and regulatory substrate from which clinical applications emerge. The principal sensor modalities ECG (user-initiated Lead I-equivalent on smartwatches, or continuous via patch), photoplethysmography (heart rate, HRV, SpO2, and via pulse transit time a BP correlate), accelerometry (step count, activity intensity, sleep staging, fall detection), and remote dielectric sensing (pulmonary fluid estimation in HF) are described with technical precision. Emerging modalities including phonocardiography and ballistocardiography remain largely confined to research settings.

The regulatory discussion is particularly well handled. The authors map CE and FDA classification systems in parallel, clarifying a distinction critical for clinical practice: regulatory approval is feature-specific, not device-wide. The Apple Watch ECG, for instance, excludes AF detection between 100–150 bpm and limits ECG output to informational use only. Even when the same hardware carries both consumer and medical device designations as with the Apple Watch Series 4+, Samsung Galaxy Watch Active2+, and Withings ScanWatch clinicians must understand which features are covered and which are not. This regulatory literacy is precisely what cardiologists need when patients arrive at clinic bearing a consumer-generated alert.

Evidence across the prevention continuum

Hughes et al. [1] structure their clinical evidence around a four-tier prevention framework. At the primordial level, meta-analyses confirm that activity trackers produce modest, short-term step-count gains in both children and healthy adults, but these rarely persist without concomitant behavioural change support wearables are facilitators, not standalone behaviour-change tools. Multi-component interventions incorporating personalised feedback, adaptive goal-setting, and gamification show greater promise. In primary prevention, the NAVIGATOR study demonstrated that baseline step count independently predicted cardiovascular risk (HR 0.90 per 2,000 steps/day), while ENGAGE showed benefit of wearable-based interventions in socioeconomically disadvantaged adults with atherosclerotic risk a population critically underrepresented in most wearables research. The authors' consistent attention to this equity dimension is commendable and sets a standard the field should adopt broadly.

Atrial fibrillation detection: population scale and clinical limits

The AF evidence is the review's most data-rich domain. PPG-based screening demonstrated feasibility at population scale Apple Heart Study (N=419,297; PPV 84%) and Fitbit Heart Study (N=455,699; PPV 98%) but the AMALFI trial, the most clinically rigorous RCT, showed only a modest increase in AF diagnosis at 2.5 years and no significant reduction in stroke rate (RR 1.08; 95% CI 0.76–1.53), a sobering result the authors engage with honestly. Lowres et al. [4] deepen this analysis: consumer devices expand detection into populations who would never self-present, but simultaneously generate an "incidentalome" of unconfirmed arrhythmia alerts in low-risk individuals. Both papers converge on the same conclusion: clinical integration requires not only analytical validation but structured diagnostic and management pathways accounting for pre-test probability and downstream investigation capacity infrastructure most health systems have not yet built. 

Heart failure monitoring: from remote detection to outcome prediction

The tertiary prevention section is the review's most clinically urgent. Hughes et al. [1] cite LINK-HF — ML-based HF readmission detection at 85% specificity, with alerts preceding hospitalisation by 6.5 days and CHIEF-HF, the first fully decentralised remote RCT in HF, where daily step count correlated with quality-of-life scores and with canagliflozin treatment response. The TRUE-HF study [2] provides the decisive scientific complement: deep learning on Apple Watch accelerometry and heart rate derived daily pVO2 estimates correlating with laboratory CPET (r = 0.85); each 10% decline independently predicted a 3.62-fold increased hazard of unplanned healthcare events. Martini [3] articulates the conceptual shift: not cross-sectional screening against population thresholds, but personalised trajectory monitoring detecting deviation from an individual's own haemodynamic baseline pointing toward closed-loop interventional triggers that could avert decompensation before hospitalisation occurs.

Cuffless blood pressure monitoring: promising but not yet ready

Hughes et al. [1] correctly flag cuffless BP as technically promising but not ready for clinical use: pulse transit time methods require cuff calibration, are susceptible to posture and ambient temperature, and BP variability though prognostically relevant cannot yet be reliably quantified by consumer wearables. Routine clinical use is not currently recommended. Choi et al. [5] advance this frontier with methodological rigour, validating a Samsung Galaxy Watch PPG algorithm against invasive continuous central arterial pressure the most demanding available reference standard. The device detected mean arterial pressure changes within under one second, substantially faster than prior wearable BP systems. The authors are explicit, however, that their protocol did not satisfy current validation requirements and that ambulatory data in diverse populations are absent. The gap between this technically impressive proof-of-concept and a clinically deployable tool for the 1.3 billion individuals with hypertension worldwide remains substantial. The consequences of deploying under-validated cuffless monitors at scale false reassurance in undertreated hypertensives are measurable in stroke and myocardial infarction rates.

Artificial intelligence, equity, and the implementation agenda

Hughes et al. [1] distinguish signal processing AI (artefact rejection, physiological estimation from sensor data) from clinical decision support AI (transforming processed signals into diagnostic or prognostic outputs), exemplified by the TRUE-HF deep learning model [2]. The equity dimension is sharply framed: models trained on non-representative populations risk producing degraded performance in the demographic subgroups most disadvantaged by conventional healthcare a risk that must be addressed in study design, not post hoc. The authors' implementation agenda is tripartite: integration of wearable data into EHR systems, development of standardised evidence-based clinical workflows, and clinician training programmes. The absence of all three currently represents the binding constraint on clinical translation. Alert fatigue, false-positive liability, and the lack of decision-support architecture mean that even analytically validated wearable signals risk being ignored or misacted upon in routine practice.

Conclusions : technical maturity outpacing clinical infrastructure

Hughes et al. [1] offer a landmark framework at a genuine inflection point in wearable cardiovascular medicine. The review's architecture from sensor technology and regulation through the prevention continuum to AI and clinical integration provides a reference that researchers, clinicians, and health system leaders can act upon. Read alongside Lowres et al. [4] on the AF detection pathway challenge, Choi et al. [5] on the technical frontier of cuffless BP, and Gao et al. [2] with Martini [3] on HF outcome prediction, a coherent and demanding picture emerges: technical capability is now maturing faster than clinical infrastructure. Wearables can detect arrhythmias, track haemodynamic decline, and predict hospitalisations. The work that remains is generating randomised outcomes evidence in diverse populations, building governance frameworks for AI, integrating wearable data into EHRs, and designing the clinical pathways that will ensure this technology reaches and benefits those who carry the greatest burden of cardiovascular disease.