Metabolic Aging Clocks Built From 18,000 People Predict Disease and Death Risk

Scientists in China have assembled one of the most detailed portraits yet of how the human body ages at the molecular level, and the result is a set of predictive tools that could reshape how clinicians assess chronic disease risk. By analyzing metabolic fingerprints from more than 18,000 individuals spanning nearly a century of life, researchers constructed mathematical models of biological aging that outperform simple chronological age in forecasting conditions such as fatty liver disease, hypertension, diabetes, and even death from any cause. The study, published in Science Translational Medicine, offers what its authors describe as a scalable and clinically interpretable framework for disease prevention, early detection, and the study of healthy aging across the entire human lifespan.

The research, led by Linlin Cao and colleagues, drew on plasma samples collected from 18,476 participants between one and 96 years of age at six clinical centers across China. Rather than focusing on a narrow age band or a single disease population, the team deliberately captured the full arc of human development, from childhood through extreme old age. Each participant’s blood plasma was subjected to metabolomic profiling, a technique that measures the small molecules, including lipids, amino acids, and other intermediates of metabolism, that circulate in the bloodstream and reflect the combined output of genes, environment, diet, and organ function. Because metabolites sit at the functional endpoint of biology, they are often more sensitive to the real-time physiological state of the body than genetic sequences, which remain fixed throughout life.

From this vast dataset, the researchers identified 56 metabolites whose abundance shifted in patterns reliably tied to age. These were not random fluctuations. The molecules formed coherent signatures of the biochemical remodeling that accompanies aging, capturing changes in energy metabolism, inflammatory pathways, and lipid handling that accumulate as tissues and organs gradually lose resilience. The team then used ensemble learning, a machine learning strategy that combines multiple predictive models to improve accuracy and robustness, to distill these signatures into quantitative aging clocks. The approach mirrors the logic of epigenetic clocks, which estimate biological age from DNA methylation patterns, but operates instead on the metabolic layer, which is cheaper to measure in clinical settings and more directly tied to current health status.

The central product of this modeling effort is a metric the researchers call metabolomic age acceleration, abbreviated mAA. The concept is elegantly simple: the models estimate a person’s metabolic age from their plasma profile, and that estimate is compared with the person’s actual chronological age. If the metabolic age exceeds the chronological age, the individual is said to exhibit age acceleration, a signal that their body is aging faster at the biochemical level than the calendar would suggest. Conversely, a metabolic age below chronological age points to slower biological aging. What elevates the finding from an interesting curiosity to a clinically meaningful measure is its predictive power. In the cohort, higher mAA correlated with an elevated risk of developing metabolic-associated fatty liver disease, hypertension, and type 2 diabetes, as well as with a greater likelihood of death from any cause as participants were followed over time.

These associations matter because they suggest that metabolic aging clocks capture something real about the trajectory of health rather than merely restating how old someone is. Chronological age is already one of the strongest risk factors for virtually every chronic disease, yet two people of the same age can occupy dramatically different positions on the path to illness. A biomarker that quantifies that divergence, using a single blood draw and a standardized computational pipeline, could allow physicians to identify high-risk individuals years before symptoms appear, when lifestyle interventions and preventive treatment are most effective. The authors emphasize that the framework is designed to be scalable, meaning it could in principle be deployed across large health systems and population screening programs without the prohibitive costs associated with whole-genome sequencing or repeated imaging.

The study did not stop at the general population. The team also examined a separate disease cohort, and in doing so uncovered a second metabolic metric they term age deviation. Unlike mAA, which is calibrated on healthy aging trajectories, age deviation quantifies how far an individual’s metabolic profile strays from the expected pattern for their age. In patients, higher age deviation was associated with more severe and more systemic disease, suggesting that the measure registers the degree to which illness disrupts the body’s normal metabolic architecture. This distinction is conceptually important. Age acceleration may reflect the slow drift of biological aging, while age deviation appears to capture acute or pathological departures from that drift, offering a complementary lens on the relationship between aging and disease.

One of the study’s most valuable contributions is the sheer depth of the underlying data. Metabolomic studies of aging have often been limited to small cohorts, restricted age ranges, or single measurement platforms, making it difficult to distinguish genuine aging signals from noise, cohort effects, or confounding by disease. By profiling nearly 18,500 individuals across the full lifespan, the Chinese team produced a reference map of how hundreds of circulating metabolites change from infancy to the tenth decade of life. This treasure trove, as the release describes it, will serve as a resource for other researchers seeking to understand which biochemical pathways drive aging, which are merely passengers, and how environmental factors such as diet, region, and socioeconomic conditions modify the aging process in different populations.

The clinical implications extend in several directions. First, the mAA metric could be incorporated into routine health checkups as a single interpretable number, analogous to how blood pressure or cholesterol are used today, flagging individuals whose metabolic systems are aging prematurely. Second, the age deviation measure could serve as a severity marker in hospitalized patients or those with chronic conditions, helping clinicians gauge how systemically a disease has disturbed normal physiology. Third, the clocks could accelerate clinical trials of interventions intended to slow aging, from exercise and dietary programs to future geroprotective drugs, by providing a quantitative endpoint that responds over months or years rather than waiting decades for hard outcomes such as mortality. The researchers have made their age prediction framework publicly accessible through a web interface hosted by their laboratory, lowering the barrier for other scientists to apply and validate the tools.

As with any biomarker study, important questions remain before metabolic aging clocks enter everyday practice. The cohort was drawn exclusively from Chinese clinical centers, and metabolite levels can vary with ancestry, diet, and local environmental exposures, so validation in other populations will be essential. The associations between mAA and disease risk, while robust within the study, will need to be tested prospectively to determine whether the metric genuinely adds predictive value beyond established clinical risk scores. It is also not yet clear whether interventions that lower metabolic age acceleration translate into longer, healthier lives, a question that only longitudinal trials can answer. Nevertheless, the scale and rigor of this study mark a significant advance in the effort to quantify aging as a measurable, modifiable process, and they support the authors’ conclusion that plasma metabolomics offers a practical foundation for studying disease prevention, early detection, and healthy aging across the life course.

The work arrives at a moment of surging interest in geroscience, the field devoted to understanding and intervening in the biology of aging. Epigenetic clocks have already demonstrated that biological age can be estimated from molecular data and that it varies meaningfully between individuals, but their reliance on DNA methylation arrays has limited clinical uptake. Metabolic clocks, built from standard blood chemistry and machine learning, could prove more adaptable to existing laboratory infrastructure. If subsequent studies confirm the predictive power reported here, a routine blood test may one day tell patients not just how old they are, but how fast they are aging, and whether the trajectory they are on leads toward chronic disease or toward a longer healthspan. For now, the study stands as a demonstration that the molecular clocks of human aging can be read from a vial of blood, and that what they reveal may be among the most informative numbers in preventive medicine.

Subject of Research: Metabolomic aging clocks that predict chronic disease and mortality risk across the human lifespan

Article Title: Large clinical study in China uncovers “clocks” of human aging that may predict chronic disease and mortality

Article References: Large clinical study in China uncovers “clocks” of human aging that may predict chronic disease and mortality. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: aging clocks, metabolomics, biomarkers, chronic disease, mortality prediction, fatty liver disease, hypertension, diabetes, machine learning, Science Translational Medicine, healthy aging, plasma metabolites