A new human study has found a possible link between time spent on social media and one molecular measure of faster aging. The more consequential finding, however, is how quickly that signal became uncertain when the analysis was asked harder questions.
Published August 6 in PLOS Digital Health, the study examined 323 adults from the Midlife in the United States Refresher Study. Participants had reported their social-media use during brief telephone interviews over eight consecutive days and later provided blood samples for DNA methylation analysis. The researchers compared average daily use with two epigenetic measures, DunedinPACE and GrimAge2.
The result was not a verdict that social media accelerates human aging. It was a compact demonstration of how a provocative association can narrow as researchers move from a simple model toward a more demanding one.
Two clocks, two different questions
DunedinPACE and GrimAge2 should not be treated as interchangeable readings of a single biological age. DunedinPACE was developed to estimate the current pace of physiological change from patterns of DNA methylation in blood. GrimAge2 was designed around methylation signals associated with mortality and age-related morbidity risk.
That distinction mattered. Each additional hour of reported daily social-media use was associated with a 0.19-standard-deviation higher DunedinPACE score after adjustment for demographic factors, smoking, alcohol use, physical activity, depression and anxiety. The confidence interval ran from 0.08 to 0.29 standard deviations, and the result was statistically significant.
But the study found no meaningful association with GrimAge2 in any model. A signal appeared in the pace-oriented clock, not in the cumulative risk-oriented clock. That divergence is not a technical footnote; it is the boundary of the claim.
TENS analysis: The model ladder matters
The study becomes more informative when read as a four-step stress test, not as a single headline coefficient. The first model adjusted for the interval between the diary and blood assessments plus demographic and socioeconomic factors. Later models added health behaviors and self-reported mental health. Across those steps, the DunedinPACE association remained similar.
The fourth model added body mass index. The estimated association then fell from 0.19 to 0.09 standard deviations, its confidence interval crossed zero, and its probability value rose to 0.088. Under the study’s prespecified threshold, the result was no longer statistically significant.
That does not prove BMI “explains” the relationship. BMI could sit somewhere along a behavioral or metabolic pathway, confound the association, or simply reflect other unmeasured processes shared with social-media use and methylation. Because the data are cross-sectional, the study cannot tell which interpretation is correct. It also cannot establish whether more time online preceded the methylation pattern.
This model ladder changes the practical reading. The evidence supports a preliminary association worthy of better measurement. It does not support a claim that an extra hour online causes a quantifiable increase in biological aging.
A digital exposure measured with an analog instrument
The exposure measure deserves equal scrutiny. Participants were asked how many minutes they had spent on websites such as Facebook, Twitter and MySpace. The data were collected from 2012 through 2014, when both platforms and usage patterns differed from today’s feeds, short-form video and algorithmic recommendations. Average reported use was 25.5 minutes a day, with wide variation.
Repeated diaries are stronger than a single long-range recollection, but self-reported duration still cannot reveal what participants saw, whether an interaction was supportive or hostile, or whether the time displaced sleep and movement. The analytic sample was also small, predominantly White and selected from people who completed both diary and biomarker projects.
Those limits prevent broad conclusions about younger users, current platforms or different racial and ethnic groups. They also expose a measurement mismatch: a finely quantified molecular outcome was paired with a coarse estimate of a rapidly changing digital environment.
What a decisive study would need
The next useful experiment is not another cross-sectional survey with a bigger sample. It is a longitudinal design that records objective device activity, separates passive consumption from communication, captures sleep and movement, and measures methylation more than once. Diverse recruitment would be essential, as would repeating the analysis across multiple aging biomarkers rather than promoting whichever clock produces the strongest association.
The Midlife in the United States program provides valuable infrastructure by linking social experience with biological data. The original DunedinPACE and GrimAge2 research also shows why these clocks can be useful at the population level. Yet neither metric converts an observational association into a diagnosis, a lifespan forecast or medical advice.
The study’s real contribution is a research agenda. Digital environments may belong in models of healthspan, but duration alone is an incomplete exposure and one epigenetic clock is an incomplete outcome. The responsible next step is to improve both sides of the measurement before turning a faint signal into a health claim.
Sources: PLOS Digital Health; Midlife in the United States, University of Wisconsin–Madison; eLife; Aging.
TENS Magazine conceptual illustration
