Digital Biomarkers in Longevity Research

Key Takeaways

Digital biomarkers are objective, sensor-derived measures collected through devices such as wearables, smartphones, and remote monitoring tools. [1] [2]
In longevity research, they most often capture functional and physiological domains such as activity, gait, sleep, heart rhythm, balance, and real-world mobility. [3] [6]
Their main advantage is repeated measurement in daily life rather than a single clinic snapshot. [4] [6]
Their main limitation is that device outputs are not automatically validated biomarkers; algorithms, adherence, context, and population fit strongly affect interpretation. [3] [4]

Who This Is Useful For

This page is useful for readers trying to understand how data from wearables, smartphones, and home sensors fit into longevity research. It is especially relevant for interpreting studies that use step counts, accelerometry, sleep patterns, heart-rate signals, gait features, or sensor-based frailty measures as ageing-related outcomes. [1] [3]

What Digital Biomarkers Measure

A digital biomarker is not just any number shown by an app. In biomedical research, the term generally refers to an objective, quantifiable physiological or behavioural measure collected through digital technology and linked to a defined biological, clinical, or functional construct. [1] [2] [3]

In longevity research, these measures often sit between laboratory biomarkers and daily function. A blood protein or DNA methylation pattern can capture molecular state, while a wearable accelerometer can capture movement, rest, gait rhythm, and changes in activity across normal environments. [4] [5] [6]

Common Digital Biomarker Types

Signal Type	Examples	Why Researchers Use It	Main Limitation
Activity patterns	Step counts, acceleration, intensity, sedentary time, day-night activity	Capture real-world behaviour linked to function, morbidity, and mortality risk	Influenced by occupation, season, device wear, disability, and environment
Gait and mobility	Gait speed, stride variability, walking bouts, turn patterns, balance features	Reflect neuromuscular control, frailty, and cognitive-motor ageing	Requires careful sensor placement, algorithms, and validation against functional outcomes
Cardiovascular signals	Resting heart rate, heart-rate variability, rhythm irregularity, recovery patterns	Can reflect autonomic regulation, cardiovascular reserve, and physiological stress	Shift with medication, illness, sleep, stress, training status, and sensor quality
Sleep and circadian signals	Sleep timing, sleep duration, fragmentation, rest-activity rhythm	Connect ageing research with recovery, cognition, metabolism, and circadian regulation	Consumer sleep staging is imperfect and can vary across devices and algorithms
Composite digital scores	Frailty classifiers, activity-derived age models, digital risk scores	Summarize complex sensor data into interpretable research endpoints	Model outputs depend on training data, target outcome, and population transportability

Why Researchers Are Interested

Digital biomarkers are attractive because ageing is not only molecular. Many ageing-relevant outcomes appear as changes in reserve, mobility, recovery, sleep, rhythm, and day-to-day behaviour. Sensor data can capture these features repeatedly, which may reveal variation that is missed when function is measured only during occasional clinic visits. [3] [4] [5]

Large cohort studies show why this matters. In UK Biobank, wrist-worn accelerometers enabled objective activity measurement in more than 100,000 participants, creating a scale of behavioural phenotyping that was difficult with clinic-based testing alone. [6] Subsequent work linked wearable-measured activity patterns with future health risks across prospective cohorts. [7] [8]

Where Digital Biomarkers Fit in Longevity Research

Digital biomarkers are usually best understood as functional and physiological measures rather than direct molecular clocks. They can complement epigenetic, proteomic, metabolomic, inflammatory, and imaging biomarkers by showing whether ageing-related biology is reflected in behaviour or performance. [5] [9]

This makes them useful in trials and cohort studies where researchers want outcomes that are frequent, low-burden, and closer to daily life. For example, a study may track whether an intervention changes activity rhythms, walking behaviour, or sleep regularity alongside blood-based biomarkers. That does not make the digital signal a complete ageing measure; it makes it one domain-specific readout within a broader biomarker framework. [3] [5]

Digital Age Models

Some studies build age-prediction models from sensor data. In one UK Biobank analysis, researchers used machine-learning models to predict age from week-long wrist accelerometer recordings and then studied the gap between predicted and chronological age as an activity-derived ageing measure. [10]

These models are conceptually similar to other ageing clocks in one respect: they summarize patterns associated with age in a particular dataset. Their interpretation is also similarly limited. A digital age estimate reflects the measured sensor domain and the training target; it is not a direct measurement of whole-body biological age. [5] [10]

Why Interpretation Is Hard

Digital biomarker data are sensitive to context. Step counts and acceleration patterns can vary with weather, shift work, disability, caregiving, occupation, device wear time, and local environment. Heart-rate and sleep measures can shift with medication, infection, alcohol, stress, and sensor placement. These influences are not noise only; they may be biologically relevant, but they make causal interpretation difficult. [4] [6]

Algorithms also matter. Two devices may measure the same person but produce different estimates because they use different sensors, sampling rates, preprocessing steps, and proprietary models. For that reason, digital biomarker research needs validation against defined outcomes rather than assuming that a convenient device metric is automatically meaningful. [3] [4]

Evidence Quality and Interpretation

Confidence is strong that wearable and sensor-derived measures can capture real-world behaviours and physiological patterns at a scale that conventional clinic visits cannot. Large accelerometer cohorts and methodological reviews support their value for population-level research. [4] [6]

Confidence is also strong that some digital measures, especially activity and mobility features, are associated with future health outcomes in observational cohorts. Prospective studies link wearable-measured activity with mortality and incident disease risk, although those associations do not by themselves prove causality. [7] [8]

Confidence is moderate that digital biomarkers can improve ageing and longevity trials by adding frequent, real-world functional endpoints. This use depends on selecting measures with clear biological meaning, acceptable reliability, and validation for the population being studied. [3] [5] [9]

Confidence is weaker for treating consumer device outputs or digital age scores as stand-alone clinical measures of biological age. Device differences, model dependence, adherence, and population transportability remain important constraints. [2] [3] [10]

What This Does Not Mean

It does not mean every wearable metric is a validated biomarker. A measure needs a defined construct and evidence that it tracks that construct reliably. [1] [3]
It does not mean digital biomarkers replace molecular biomarkers; they usually capture different layers of ageing biology and function. [5]
It does not mean a digital age estimate is a complete measure of biological age. It is a model output tied to a sensor domain and training dataset. [10]
It does not mean observational associations between activity and health outcomes prove that changing the metric will directly change ageing biology. [7] [8]

Practical Interpretation Examples

If a wearable shows low activity: that may indicate reduced mobility, illness, environment, or behaviour, but it is not automatically a direct ageing diagnosis. [4] [6]
If a gait sensor detects higher variability: that may reflect cognitive-motor or frailty-related risk, but interpretation depends on the protocol and comparator group. [9]
If a digital model estimates someone as older: that means their sensor pattern resembles older participants in that model, not that their whole body has been measured as older. [10]

Summary

Digital biomarkers extend longevity research into real-world behaviour and function. They can capture activity, gait, sleep, cardiovascular signals, and composite functional patterns repeatedly and at population scale. Their value is strongest when they are treated as validated, domain-specific research measures rather than as automatic indicators of whole-body biological age. [3] [5] [6]

References

FDA-NIH Biomarker Working Group. (2016). BEST (Biomarkers, EndpointS, and other Tools) Resource. https://www.ncbi.nlm.nih.gov/books/NBK326791/
Alonso, A. K. M., et al. (2024). Definitions of digital biomarkers: a systematic mapping of the biomedical literature. BMJ Health & Care Informatics. https://pmc.ncbi.nlm.nih.gov/articles/PMC11015196/
Coravos, A., Khozin, S., & Mandl, K. D. (2019). Developing and adopting safe and effective digital biomarkers to improve patient outcomes. npj Digital Medicine. https://pubmed.ncbi.nlm.nih.gov/30868107/
Hicks, J. L., et al. (2019). Best practices for analyzing large-scale health data from wearables and smartphone apps. npj Digital Medicine. https://www.nature.com/articles/s41746-019-0121-1
Moqri, M., Herzog, C., Poganik, J. R., et al. (2023). Biomarkers of aging for the identification and evaluation of longevity interventions. Cell. https://pmc.ncbi.nlm.nih.gov/articles/PMC11088934/
Doherty, A., et al. (2017). Large Scale Population Assessment of Physical Activity Using Wrist Worn Accelerometers: The UK Biobank Study. PLOS ONE. https://pmc.ncbi.nlm.nih.gov/articles/PMC5287488/
Strain, T., et al. (2020). Wearable-device-measured physical activity and future health risk. Nature Medicine. https://pmc.ncbi.nlm.nih.gov/articles/PMC7116559/
Khurshid, S., et al. (2022). Wearable accelerometer-derived physical activity and incident disease. npj Digital Medicine. https://www.nature.com/articles/s41746-022-00676-9
Zhou, H., Razjouyan, J., Halder, K., et al. (2021). Digital biomarkers of cognitive frailty: the value of detailed sensor-based gait assessment. Brain Sciences. https://pmc.ncbi.nlm.nih.gov/articles/PMC8578566/
Le Goallec, A., et al. (2023). Machine learning approaches to predict age from accelerometer records of physical activity at biobank scale. PLOS Digital Health. https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0000176

Educational Disclaimer

This content is provided for educational purposes only and does not constitute medical advice.