A consumer fitness tracker can record when a person moves. That does not mean it can read biological age. A new peer-reviewed study nevertheless suggests that long-term activity rhythms may carry a useful aging-related signal when they are paired with clinical data.
Published August 22 in Nature Communications, the study analyzed multi-year Fitbit records from 2,222 participants in the National Institutes of Health’s All of Us Research Program. Researchers linked hourly patterns of movement with annual estimates of PhenoAge, a composite derived from chronological age and routine clinical biomarkers. The dataset represented 8,447 person-years of observation.
What the study found
Stronger daily activity rhythms were associated with 26 to 46 percent lower odds of being classified as biologically older than expected. Greater day-to-day regularity was associated with 9 to 13 percent lower odds, while a later daily activity peak was associated with 22 percent higher odds. The authors also reported different timing patterns by sex, although subgroup results require additional validation.
Those figures describe associations across the observed range of the study’s rhythm measures. They do not show that making a schedule more regular will reduce anyone’s biological age by the same amount. Illness, mobility, employment, sleep, caregiving, medication and the built environment can all influence movement patterns. Faster biological aging could also weaken a person’s daily rhythm, creating reverse causation.
TENS analysis: a signal built from two proxies
The important distinction is that this system stacks two measurement layers without directly observing aging itself. Fitbit steps are a behavioral proxy for rest-activity rhythm, not a measurement of the body’s internal circadian phase. PhenoAge is a risk-oriented model built from clinical biomarkers, not a stopwatch for lifespan. Agreement between the two can identify a scalable signal, but it cannot by itself establish a mechanism, diagnosis or intervention target.
This helps explain both the study’s promise and its limits. A wearable can collect repeated observations in ordinary life for far longer than the week of monitoring common in research actigraphy. That long view can smooth over an unusual vacation, infection or deadline. Yet consumer devices also miss non-stepping activities such as cycling and resistance training, and incomplete wear can look like inactivity. Official All of Us guidance notes that Fitbit data arrive from the device platform without additional program-level cleaning and that adherence bias requires deliberate handling.
The selection funnel matters
The cohort also became much smaller as the data requirements tightened. The earlier public methods record identified 34,217 people with Fitbit data, consented health records and valid physical measurements. Of those, 11,698 had complete step data for a full year, and 2,222 had repeated PhenoAge estimates aligned with valid Fitbit years. That final group is about 6.5 percent of the starting pool and 19 percent of those who cleared the full-year activity screen. The resulting depth is valuable, but the funnel favors people able and willing to own, sync and consistently wear a device while generating sufficient clinical laboratory data.
The authors described the Fitbit cohort as disproportionately female, White, college-educated and more active than the national average. That limits immediate generalization to people with disabilities, irregular work, unstable housing, limited access to wearables or different patterns of clinical care. Missing laboratory measurements can add another layer of uncertainty to an EHR-derived age estimate.
Why one wearable score would be premature
A May study in JAMA Network Open provides a useful comparison. In 207 middle-aged and older adults wearing research-grade actigraphs, stronger and more regular rhythms were also associated with favorable results on some epigenetic clocks. But the direction and strength varied by clock, age, sex and race and ethnicity. Timing results were especially inconsistent. Together, the studies suggest that rhythm strength may be a more stable research signal than a single ideal bedtime or activity peak.
The next validation ladder has three rungs. First, independent cohorts must show that the same rhythm features survive different devices, populations and biological-age models. Second, studies must demonstrate that a wearable signal predicts meaningful outcomes beyond information already available from mobility, disease burden and routine laboratory tests. Third, randomized interventions would need to show that deliberately changing a rhythm improves health outcomes. Until then, the finding supports research-grade risk stratification, not a consumer longevity score.
No lifespan or healthspan extension was tested. The evidence is human and longitudinal, but observational. The strongest conclusion is that repeated wearable data may help researchers map aging-related risk at scale while exposing how much calibration, representation and causal validation remain unfinished.
Sources: Nature Communications; All of Us Research Program; JAMA Network Open.
TENS Magazine conceptual illustration
