A new Nature Medicine study of 60,000 people found that your body does not age as a whole. Different cell types age at completely different speeds, and the pattern predicts which diseases you will develop.
Biological aging research has long operated with a simplified model: some people age faster than others, some organs wear out before their time, but within a single individual, the body ages more or less as a system. The brain slows, the heart stiffens, the immune system weakens. These changes happen in parallel and tend to move together, driven by the same underlying biological processes of cellular damage, inflammation, and repair failure.
A new study published in Nature Medicine is the largest and most granular challenge to that model yet assembled.
Researchers led by David Furman at the Stanford University School of Medicine, in collaboration with scientists at UC San Francisco and Karolinska Institutet, analyzed blood plasma samples from 60,542 people drawn from three independent research cohorts. From each sample, they measured the levels of more than 7,000 proteins using two separate high-throughput proteomics platforms, the SomaScan aptamer array and the Olink antibody array, allowing them to cross-validate their findings independently.
Using the Human Protein Atlas, which maps which proteins are preferentially produced by specific cell types in the body, they linked each plasma protein to its likely cellular origin. This allowed them to build machine learning models that estimate not just how fast a person’s overall body is aging, but how fast each of more than 40 distinct cell types is aging separately.
The cell types in the analysis spanned an extraordinary range of biological systems: neurons, astrocytes, oligodendrocytes, microglia, various immune cell populations including T cells, B cells, macrophages and natural killer cells, endocrine cells, epithelial cells, hepatocytes, and musculoskeletal cell types. Each got its own aging clock, calibrated against chronological age and validated in independent cohorts.
The result was a picture of human aging that the standard model cannot capture: every person’s body is aging heterogeneously, with different cell types following different trajectories, and those trajectories predict disease and death with precision that single-organ or whole-body aging clocks cannot match.
What the data revealed about aging within a single person
The most striking finding in the study is not the existence of cellular aging heterogeneity, which prior smaller studies had suggested, but its prevalence and its clinical consequences.
In the full cohort of 60,542 people, 20 to 25 percent of individuals showed dramatically accelerated aging in at least one cell type relative to their chronological age. These were people whose neurons, or astrocytes, or immune cells, or hepatocytes were behaving as if they were substantially older than the rest of their body.
Between one and three percent of the cohort showed accelerated aging in ten or more cell types simultaneously, representing a pattern of broad cellular deterioration that the researchers call multi-cellular aging. These individuals showed the highest mortality risk in the follow-up analyses.
The heterogeneity was not simply noise in the data. The models were independently validated across two separate protein measurement platforms and three separate cohorts from different countries, which is an unusually rigorous standard in proteomics research. The same patterns appeared consistently regardless of which measurement technology or which population was used to build the model.
Critically, each person’s pattern of cellular aging was stable over time. In a subset of participants with longitudinal samples collected ten years apart, the relative ordering of which cell types were aging fastest within a given individual remained largely consistent. The person whose astrocytes were aging faster than their neurons at time one tended to show the same relative pattern at time two. Cellular aging heterogeneity is not random noise. It is an individual signature.
How the pattern predicts disease fifteen years later
The study followed participants for up to 15 years after their initial blood sample, tracking which diseases developed and when. The cellular aging clocks were then tested as predictors of those outcomes.
The results were specific in ways that a single whole-body aging clock cannot be. Different cell type aging signatures predicted different diseases, and the relationships made biological sense.
Accelerated heart cell aging predicted heart failure. Accelerated lung cell aging predicted chronic obstructive pulmonary disease. Accelerated kidney cell aging predicted hypertension and renal disease. Accelerated liver cell aging predicted metabolic liver conditions. Each organ’s cellular aging clock was most predictive of the disease associated with that organ.
For Alzheimer’s disease, the most powerful predictor was not overall brain aging but the specific aging of astrocytes, the star-shaped support cells that maintain neuronal function, regulate the blood-brain barrier, and clear metabolic waste from the brain. People with the oldest astrocytes relative to their chronological age showed substantially higher Alzheimer’s risk in the 15-year follow-up.
“We show that aging is asynchronous across different cell types, meaning different cells in the same individual age at different rates,” said Furman. “Moreover, we show that this cellular aging heterogeneity is clinically relevant: accelerated aging in specific cell types predicts the development of corresponding diseases.”
The Alzheimer’s gene paradox that nobody expected
One of the study’s most counterintuitive findings concerns APOE4, the genetic variant most strongly associated with sporadic Alzheimer’s disease risk. Carriers of the APOE4 allele face significantly elevated lifetime risk of Alzheimer’s, and the biological mechanisms through which APOE4 raises that risk have been extensively studied.
When the researchers examined how APOE genotype affected cellular aging profiles, they found a paradox.
People carrying the APOE4 allele showed older astrocytes than APOE3 carriers, consistent with the elevated Alzheimer’s risk associated with the variant. But they also showed younger macrophages, the immune cells responsible for clearing cellular debris and managing inflammation throughout the body.
The APOE2 allele, which is associated with lower Alzheimer’s risk, showed precisely the opposite pattern: younger astrocytes but older macrophages.
This means the same genetic variant that accelerates aging in one cell type simultaneously appears to slow aging in another. APOE4 does not simply cause faster aging universally. It reorganizes the distribution of aging across cell types in a way that happens to increase Alzheimer’s risk specifically, through astrocyte aging, while providing some compensatory benefit elsewhere.
This finding suggests that interventions targeting APOE4’s effects may need to account for its opposing influences on different cell populations. Slowing astrocyte aging in APOE4 carriers may be the specific lever that matters for Alzheimer’s prevention, rather than any broader anti-aging effect.
What the most extreme astrocyte aging means for Alzheimer’s risk
Among the quantitative findings in the study, one stands out for its magnitude.
In individuals carrying two copies of the APOE4 allele, which represents the highest genetic risk group for Alzheimer’s disease, those who also showed extreme astrocyte aging had their Alzheimer’s risk tripled compared to APOE4 carriers with more youthful astrocytes.
Conversely, APOE4 homozygotes whose astrocytes were aging slowly showed substantially reduced Alzheimer’s risk despite carrying the highest-risk genotype. The cellular aging measurement was providing information about Alzheimer’s risk that the genetic data alone could not.
This is clinically significant because genetic testing for APOE4 has become increasingly common, and carriers frequently ask clinicians what their result means for their future. The standard answer has been that APOE4 increases risk substantially but cannot predict with certainty whether any individual will develop Alzheimer’s. The addition of astrocyte aging data from a blood proteomics test could sharpen that prediction considerably, identifying which high-risk carriers are on the fastest trajectory and which, despite their genotype, are aging more slowly in the brain cells most critical to Alzheimer’s pathology.
How lifestyle factors shape cellular aging patterns
The study examined how lifestyle and environmental factors influenced cellular aging across different cell populations, finding patterns that add nuance to the standard advice about healthy aging.
Smoking showed its effects concentrated in lung and airway epithelial cell aging, as expected, but also produced measurable acceleration in kidney, stomach, and intestinal cell aging. The biological damage of smoking is not confined to the lungs.
Exercise was associated with younger muscle cell aging, consistent with the established literature on physical activity and muscle health. More surprisingly, exercise was also associated with younger immune cell aging, consistent with research showing that regular physical activity has broad immunomodulatory effects.
Diet and body mass index showed effects across metabolic cell types including hepatocytes and adipocytes. Medications produced their own signatures: statins were associated with slower heart and vascular aging in users, consistent with their cardioprotective effects, while certain immunosuppressants showed complex patterns of cell-type-specific aging that were not uniformly beneficial or harmful.
The cell-type specificity of these environmental effects suggests that the popular framing of lifestyle as globally beneficial or harmful oversimplifies the picture. Exercise slows muscle and immune aging preferentially. Smoking accelerates lung and kidney aging preferentially. The body is not one system responding uniformly to its environment.
What a blood test revealing your cellular aging profile could mean
The most practically significant aspect of the study is that all of its findings rest on a standard blood draw. Plasma proteomics does not require tissue biopsies or organ-specific imaging. A sample of blood contains protein signals from cell types throughout the body, and those signals can be decoded to reveal how fast each cell type is aging.
This means the cellular aging profiles described in this study are, in principle, measurable today, with existing technology, in a clinical setting. The SomaScan and Olink platforms used in this research are available at specialized research and clinical labs. The limiting factors are cost, access, and the clinical evidence base needed to establish what specific cellular aging profiles should trigger clinical intervention.
The study provides the evidence base for the prediction side of that equation. It demonstrates that cellular aging profiles measured from blood predict disease and death over 15 years with specificity that current clinical tools cannot match. What remains to be established is whether any intervention, identified on the basis of cellular aging patterns, can slow the specific cell types at risk and reduce the disease outcomes that follow.
The researchers are explicit that the study is a step toward personalized aging medicine rather than a finished clinical tool. But the distance between a research finding and a clinical application has been shortening rapidly in the aging field.
“We hope this work can ultimately help physicians identify which patients are at highest risk for specific aging-related diseases before those diseases become manifest,” Furman said. “The goal is to intervene earlier, when there is still time to make a difference.”
What the study cannot establish
The cohorts were predominantly older and Caucasian, which limits how well the findings generalize to younger adults and to populations with different genetic backgrounds, environmental exposures, and disease histories. The researchers explicitly identify broader validation as essential.
The protein-to-cell-type assignments depend on the Human Protein Atlas, which links proteins to their likely cellular origin based on gene expression patterns. Some proteins are produced by multiple cell types under different conditions, which introduces uncertainty into which cell’s aging is actually being measured.
The 15-year follow-up is observational. The study shows that cellular aging profiles predict disease development, but it cannot establish that the cellular aging caused the disease or that slowing the cellular aging would prevent it. Those causal questions require intervention studies that have not yet been conducted.
What the study establishes, at a scale and resolution that justifies the description as the largest cellular aging study ever conducted in humans, is that the body ages heterogeneously, that the heterogeneity is stable and individually characteristic, and that it predicts disease and mortality with organ-specific precision. The body does not age as one thing. It ages as many things simultaneously, each at its own speed, each carrying its own information about the future.
The study, “Plasma proteomic signatures of cellular aging predict human disease”, was authored by Douglas Yao, Yingxuan Chen, Thomas W. Chu, and colleagues at Stanford University, UC San Francisco, and Karolinska Institutet, and published June 18, 2026 in Nature Medicine.
Source: Stanford University School of Medicine. DOI: 10.1038/s41591-026-04446-y
