A blood-based aging score carries information about the cells in the sample as well as the molecular marks being measured. A Genome Medicine study published October 2 makes that distinction more precise: blood-cell composition helps explain variation in epigenetic clocks, yet accounts for comparatively little of their association with later health outcomes.
For longevity research, those findings belong together. TENS Magazine’s analysis is that a useful predictor and an interpretable biological measurement require different evidence. Explaining part of a score does not automatically explain its predictive value. Equally, a score that retains predictive value has not thereby become a target whose alteration will improve health.
The mixture and the forecast
Thomas Jonkman and colleagues analyzed 4,058 human blood samples and examined associations with 176 incident health outcomes in another 18,859 people. Cell composition explained up to 53 percent of variation in DNA-methylation age, compared with up to 21 percent in age acceleration. Adjustment for composition only modestly weakened the latter’s health-outcome associations.
These percentages describe different statistical questions; they are not estimates of how much human aging is caused by particular cells. The research used observational human data and computational mixtures assembled from cell-specific methylation profiles. Those mixtures test measurement behavior, not rejuvenation in living people.
Our interpretation is that a research report should keep two accounts: what produces variation in the measurement, and what connects that measurement with an outcome. A large entry in the first account need not be a large entry in the second. Readers should be able to see both before a biological explanation is offered.
Consider a hypothetical instrument that captures several overlapping signals. Removing a conspicuous component might change many readings while leaving much of the instrument’s forecasting ability intact. That would justify investigating the remaining information. It would not identify one remaining component as the cause of the forecast, or establish what would happen if someone deliberately changed it.
What the clock was built to estimate
Earlier primary research helps explain why one label can conceal different measurement tasks. In their 2022 eLife paper, Daniel Belsky and colleagues developed DunedinPACE using changes in 19 indicators of organ-system integrity across four assessments over two decades. They then trained a methylation-based measure and evaluated it in five additional datasets.
That design aims to estimate pace from a blood sample, rather than simply reproduce a birthday. Its reported associations with morbidity, disability and mortality support its value as a research biomarker. They do not make every later use of the algorithm an independently validated test of treatment benefit.
TENS would therefore attach a short specification to any clock result: the intended quantity, the population studied, and the comparison being made. An age estimate, a deviation from an age expectation and an estimated pace should not share an unlabeled results column. Clear units are part of the scientific argument, not cosmetic presentation.
An intervention adds a separate question
The CALERIE analysis published in Nature Aging in 2023 provides a useful contrast. The parent trial randomized 220 adults without obesity to caloric restriction or a control diet for two years. Its post hoc methylation analysis found a small effect on DunedinPACE but no significant changes in PhenoAge or GrimAge estimates. The authors called for longer follow-up of disease and mortality outcomes.
This comparison is methodological, not dietary advice. The new composition study asks how a signal is assembled; CALERIE asks whether assignment to an intervention changes particular readings. Neither design permits a reader to import the other’s answer. In particular, composition sensitivity alone cannot establish why a score moved in a separate trial.
A stronger future study would state its interpretation before inspecting the results. If a biomarker shifts, investigators could examine whether accompanying cellular measurements support the proposed explanation, while reporting direct health outcomes separately. If the measurements disagree, preserving that disagreement would be more informative than selecting whichever number tells the most attractive story.
Keeping clinical claims proportional
The US Food and Drug Administration distinguishes biological indicators from surrogate endpoints used in place of clinical outcomes. Validation requires evidence connecting the surrogate with benefit in the relevant setting. FDA also cautions that even validated surrogates can miss other effects that alter a product’s overall balance of benefit and harm.
That framework suggests a practical editorial boundary: explain measurement advances fully, but keep conclusions within the experiment performed. The studies discussed here do not establish that manipulating a clock score extends human lifespan or healthspan. The October paper supplies a more specific reason to investigate the contents of a blood-based signal, without dismissing the information that survives that investigation.
TENS Magazine conceptual illustration