The rapid advancement of wearable technology and digital health drives the latest phase in the evolution of HF trial endpoints. In Table 1, examples of wearable-derived endpoints in HF populations are summarised, indicating the diversity of wearables and objectives. Wrist-worn trackers, ranging from activity trackers (dedicated fitness bands and pedometers) and smartwatches to patch-based sensors, can continuously and remotely monitor various physiologic and behavioural parameters of HF patients in their everyday environments (Fig. 1). This capability opens new possibilities for endpoint assessment in HF RCTs, as investigators can now capture patient activity levels, heart rate and rhythm, blood pressure, sleep patterns, and even cardiopulmonary metrics in real-time. Wearable sensors thus enable a shift from intermittent, clinic-based measurements to continuous monitoring, potentially providing a more granular and holistic view of how HF patients respond to therapy. For example, rather than relying only on clinic assessments of functional class, such as the NYHA classification, a trial could use daily step count or physical activity duration (derived from accelerometers) as an endpoint to quantify improvement in a patient’s functional capacity in real-world settings, as has been done with several different accelerometer devices (see examples in Table 1). Early studies have demonstrated that such wearable metrics correlate with traditional functional measures: in the recent DETERMINE-HF and EMPIRE-HF trials, baseline daily activity level tracked by an accelerometer showed a modest correlation with 6MWT distance and KCCQ scores [33,34,35]. Interestingly, changes in activity did not perfectly mirror changes in 6MWT or KCCQ, suggesting that wearable data provide complementary information about patient status [33]. This indicates that wearable-derived endpoints could enrich our understanding of treatment effects by capturing aspects of patient function that clinic tests and questionnaires might miss (for instance, spontaneous daily mobility vs. structured test performance).

Overview of Wearable Integration in Randomised Controlled Trials (RCTs). The figure illustrates a generalised workflow for incorporating wearable technology in RCTs, encompassing (I) the study workflow from patient selection to endpoint definition, (II) examples of common wearable devices and their collected physiological data, and (III) the data processing workflow from initial data collection to clinical endpoint classification

Wearables can also detect clinically important events that might otherwise go unreported between trial visits. A pertinent example in CV research is the use of smartwatch photoplethysmography to intermittently screen for atrial fibrillation (AF), as evidenced by the Apple Heart Study, which, in a virtual, app-based study, enrolled over 400,000 participants nationwide and successfully identified new AF cases (n = 153 new AF cases) [36]. In HF, continuous wearables monitoring could similarly detect subclinical fluid retention episodes or arrhythmias that portend HF decompensation; however, testing in large-scale clinical trials is needed to establish the clinical utility of such continuous monitoring [37, 38]. As an illustration, a multi-sensor wearable could track a patient’s resting heart rate, respiratory rate, and activity; an upward drift in resting heart rate coupled with reduced daily steps might signal worsening HF before the patient requires urgent care. Although invasive, devices such as the CardioMEMS system offer proof of principle that proactive detection of fluid retention can significantly impact patient management and a similar concept could be utilised in the wearables field. The CardioMEMS device measures pulmonary artery pressure directly and transmits data remotely to healthcare providers, allowing them to detect fluid overload early and intervene before clinical symptoms emerge or worsen [39, 40]. As a non-invasive alternative to CARDIOMEMS, the CardioTag system is a recent wearable device that may be used to measure right-sided heart pressures, potentially increasing accessibility and improving cost-effectiveness of hemodynamic-guided HF-management [41]. Capturing such dynamic endpoints (e.g., day-to-day variability in physiological markers or activity) enables a more nuanced evaluation of how an intervention stabilises HF trajectory or reduces volatility in a patient’s condition.

The extensive datasets generated by wearable devices constitute a promising resource within healthcare research, particularly when harnessed through artificial intelligence (AI) and machine learning (ML). These novel methodologies enable sophisticated predictive modelling, transforming continuous physiological data streams into actionable clinical insights. ML is uniquely suited to leverage these rich, dynamic datasets, facilitating the identification of subtle patterns and predictive signatures that conventional statistical methodologies might overlook [42]. Consequently, ML-driven analysis allows for nuanced risk stratification, precise phenotypic characterisation of diseases, and detailed mapping of disease trajectories. Thus, the predictive modelling capabilities inherent in ML can potentially redefine the traditional approach to primary outcomes in clinical trials, enhancing the sensitivity and accuracy of therapeutic efficacy assessments and offering novel perspectives on disease progression and patient responsiveness to interventions. For example, in large clinical trials where data collection has included continuous monitoring with a wearable device, the application of data analysis with AI and ML may potentially identify both new phenotypes of patients, new wearable endpoints, and differences in responsiveness to an intervention at different stages of disease.

Another promising aspect of wearables is their potential to facilitate decentralised trial designs. By allowing remote data capture, wearables reduce the reliance on frequent in-person visits for endpoint assessment. This can ease participant burden and improve accessibility, particularly for individuals with limited mobility or those residing far from research centres. Digital health initiatives, such as the MyHeart Counts study, have demonstrated the feasibility of remote recruitment and data collection at scale [43]. In HF RCTs, such approaches may enhance the inclusion of underrepresented populations and enable more pragmatic trial designs. Moreover, continuous out-of-clinic monitoring mitigates recall bias and yields objective, real-time data, improving dataset completeness and capturing clinically relevant changes that might otherwise go undetected.

Therefore, wearable-derived endpoints offer several opportunities: more patient-centred assessments, more sensitive detection of changes or events, and more inclusive trial conduct through remote participation.