Repeated blood measurements in 335 women reveal why the average aging curve can miss an individual’s biological path.
Two people can share a birthday without sharing the same biological journey. A new study offers a detailed look at that familiar idea: molecular changes associated with aging can move in different directions in different people, even when researchers follow them across the same period of life.
In research published in Science, a team involving King’s College London repeatedly measured gene activity and blood metabolites in 335 women from the TwinsUK cohort. Follow-up extended to eight years. The results describe shared biological patterns, but also substantial variation that a population average can conceal.
Evidence level: Early Research. This is a longitudinal human observational study, not a trial of an anti-aging treatment or a validation of a commercial biological-age test.
Watching the same people change
Much aging research compares younger and older people at one moment. That approach can identify differences between age groups, but it cannot show exactly how each person arrived at their current biological state. Differences in upbringing, health, medication or environment may also separate the groups.
The new MultiMuTHER study instead followed the same participants across at least three clinical visits. Women ranged from 32 to 80 years old at baseline, with a median age of 61. Samples were collected between 2009 and 2017; the median interval between first and last visits was six years, within the eight-year maximum.
The team studied gene expression, meaning how actively cells use particular genetic instructions, through RNA measurements in whole blood. They also examined metabolites: small molecules involved in metabolism, including products of energy use and substances influenced by food or environmental exposure. These measurements describe aspects of biological activity, not changes to a person’s underlying DNA sequence.
Averages can hide opposing directions
Across the analyses, the researchers identified longitudinal changes involving 5,061 genes and 181 metabolites. Those totals combine features with a consistent population-level direction and others whose changes were primarily individual.
Among the metabolites, 45 showed a consistent direction of change across the population. Another 136 showed more varied trajectories: a molecule might rise in some participants and fall in others, weakening or canceling the average trend. Even for genes with strong group-level patterns, individual trajectories did not always follow the same direction.
Think of a room in which half the people turn up their thermostats and the other half turn them down. The average setting may barely move, although many individuals have made a substantial change. That is an illustration of the statistical problem, not a suggestion that the body operates like a single thermostat.
Changes involved biological pathways relevant to immunity, metabolism and age-associated disease. But identifying a pathway associated with a disease is not the same as diagnosing that disease or predicting that a particular participant will develop it.
A blood sample also captures context
The study also examined genetic influences, immune-cell differences, season and the time of day samples were taken. Many molecular measurements varied with these cyclical factors. That matters because blood is not a static record of aging: it contains cells and molecules responding to ongoing physiological demands.
Environmental signals complicated the picture too. Blood levels of two PFAS pollutants declined over the study period. A change occurring as someone gets older therefore need not be caused by aging itself. It can also reflect changing exposure over calendar time.
What this means for biological-age testing
Biological-age tools try to summarize aspects of health or aging from measurable features. The new findings do not make every such tool invalid. They do suggest that reducing a complex, changing molecular profile to one number can leave important information out.
For personalized longevity medicine, the useful question may sometimes be how someone’s measurements are changing relative to their own earlier values, alongside established clinical information. Before that approach becomes a reliable service, researchers must show which changes predict meaningful outcomes and whether acting on them improves care.
A rising biomarker is not automatically harmful, and a falling one is not automatically rejuvenation. This study did not assign treatments, establish an ideal molecular trajectory or demonstrate that repeated testing extends life. It also did not prove that every individual needs a different intervention.
Important limits
All participants were women drawn from a UK twin cohort. The findings need testing in men and more diverse populations. Blood measurements cannot represent every organ, and a handful of visits cannot capture every fluctuation between samples. Statistical adjustment reduces some confounding but cannot make an observational study establish causation.
For readers, the practical takeaway is caution about treating one molecular score as a complete verdict on health. The research supports a richer understanding of variation, not a new shopping list of tests or supplements.
Why it matters
Population averages help researchers find patterns. Following individuals helps reveal what those patterns miss. Both perspectives will be needed to distinguish ordinary biological variation from changes that genuinely signal future illness.
Bottom line
Aging appears to involve shared processes with highly individual molecular trajectories. This study maps that complexity more clearly, but it does not yet tell people how to measure, slow or reverse their own aging.
Sources: El-Sayed Moustafa and colleagues, Science, published September 3, 2026; King’s College London study record and accepted manuscript. Featured image: original AI-generated conceptual illustration, not study participants or measured data.