A UK Biobank Machine Learning Map of How Physical Health Tracks the Aging Brain
Researchers at the University of Otago trained machine learning models on more than 30,000 UK Biobank participants to ask a simple question: how well can your body predict your mind? A composite panel of 19 body phenotypes (317 individual measures spanning body composition, bone, cardiovascular, lung, kidney, liver, immune, metabolic, and muscle systems) predicted general cognitive ability at a correlation of r = 0.40, explaining about 16% of the variance in cognition in people the model had never seen. Body composition and bone health were the strongest physical predictors, rivalling two of three brain imaging approaches. Critically, roughly 85% of the body–cognition signal was mirrored in brain scans, and body plus brain together captured almost all (96.8%) of the link between cognition and age. This is a correlational, cross-sectional study, not an intervention.
For centuries the idea of “a healthy mind in a healthy body” has been more proverb than measurement. This study tries to put a number on it. Using the UK Biobank, one of the largest datasets pairing cognitive testing, whole-body physiology, and brain imaging, the Otago team built machine learning models that predict a person’s general cognitive ability (a latent “g-factor” pulled from twelve cognitive tests) from their physical health alone.
The big idea is integration. Most previous work looked at one system at a time, asking whether blood pressure, or lung function, or muscle mass tracks with thinking. Here the researchers threw 19 body phenotypes into a single stacked model and asked how much of cognition the body as a whole can forecast in strangers. The answer: a correlation of r = 0.40, or about 16% of cognitive variance, tested out of sample so the number is not inflated by overfitting. Body composition measured by bioimpedance and bone mineral density measured by DXA carried most of the weight, each on its own out-predicting resting-state and structural brain imaging.
The more striking finding is the overlap. When the team compared the body signal to brain scans, about 85% of the body–cognition relationship was already visible in the brain, especially in white matter, the wiring that connects regions. In other words, the body and the brain are not telling two different stories about cognition; they are largely telling the same one. Extend this to aging and the convergence gets sharper still: brain and body markers together accounted for 96.8% of the relationship between cognition and chronological age, with 71.7% of that shared jointly.
The takeaway the authors push is a geroscience one. Cognitive aging is not a brain-only event sealed inside the skull. It is a whole-body process, and the physical systems most tied to it (adiposity, muscle quality, bone, lung capacity, glucose control, blood pressure) are the same ones clinicians already know how to monitor and, in principle, modify. The caution the authors also push, repeatedly, is that none of this proves causation.
Actionable Insights
The practical signal points to a familiar but now quantified set of levers. Because this is correlational, treat these as targets worth tracking, not proven cures.
What it found are strengths of connection between physical traits and cognitive ability across tens of thousands of people. The strength of each connection is scored from 0 (no relationship) to 1 (a perfect one-to-one match). In this kind of research, a score around 0.1 is weak, 0.3 is moderate, and 0.4 or higher is considered strong. A minus sign just means the trait goes the other way (more of it lines up with lower cognitive scores). Think of these as clues about which parts of physical health travel alongside a sharper mind, not as a to-do list with guaranteed payoffs.
The single strongest clue was muscle quality. Fat marbled inside the thigh muscles had the strongest negative link of any measure (about 0.44), meaning people with more fat mixed into their muscle tended to score lower. Higher lean, fat-free muscle mass went the opposite, healthier direction. Fat stored in the wrong places also lined up with lower scores, including fat in and around the pancreas (about 0.32) and deep belly (visceral) fat (roughly 0.27).
Blood sugar and blood pressure told the same story. Higher blood pressure (about 0.33) and higher long-term blood sugar, measured as HbA1c (about 0.25), both tracked with lower cognitive scores, as did a thicker, more clogged carotid artery wall (about 0.25).
On the positive side, better lung capacity was one of the strongest healthy signals (breathing power, FEV1, about 0.40), along with height and bone strength, and a growth-related hormone called IGF-1 (about 0.27).
The practical read: the physical traits most tightly linked to a healthier aging brain are keeping muscle strong and low in fat, avoiding excess belly and organ fat, and keeping blood pressure and blood sugar in check. Taken all together, a person’s full physical profile predicted their cognitive ability about as well as some brain scans did, explaining roughly one sixth of the differences between people. That is a meaningful chunk, but it also means most of what shapes cognition lies elsewhere.
Context and Source
- Open Access Paper: Exploring the Link Between Body Physiology and Cognition: The Role of the Brain and Aging.
- Authors and institution: Irina Buianova and Narun Pat, Department of Psychology, University of Otago, Dunedin, New Zealand.
- Journal: npj Aging (Nature Portfolio, Springer Nature). ISSN 2731-6068.
- Impact evaluation: The impact score of this journal is 13.0 (2025 Journal Impact Factor, per Nature’s official metrics page; 5-year JIF 12.2), evaluated against a typical high-end range of 0 to 60+ for top general science, therefore this is a Medium-to-High impact journal. It is a Q1 specialist aging journal, well regarded within geroscience