UK Biobank is a British cohort of about 500,000 volunteers aged 40 to 69, recruited between 2006 and 2010 and followed for at least thirty years. Genomic data, imaging, health records and metabolomic profiles are linked together. Most of the biomarkers and risk models used in personalized health were built or validated on this resource.
What is UK Biobank?
UK Biobank is both a biobank and a prospective cohort: between 2006 and 2010, half a million British adults agreed to answer questionnaires, undergo physical measurements and give blood, urine and saliva samples, and then to be followed for decades through health system data. The goal is to understand why some people develop a disease and others do not.
The project was funded by the Medical Research Council, the Wellcome Trust, the UK government and charities; it is based in Stockport, in Greater Manchester. Since March 2012, the database has been open to researchers worldwide, from both the public and private sectors, upon application and for health research projects. As of November 2023, more than 9,000 peer-reviewed publications drew on its data.
Why does a prospective cohort change everything?
In a prospective cohort, measurements are taken first, in healthy people, and then researchers wait. Ten or fifteen years later, they know who has had a heart attack, diabetes or dementia, and they can go back to the baseline measurements to identify what already set these people apart. This is how a biomarker is shown to be associated with a future disease, and how strongly. Such a study reveals associations; it does not prove that a marker causes the disease, nor that acting on it prevents the disease.
What data has UK Biobank collected?
For the same individuals, UK Biobank combines lifestyle data, clinical measurements, the whole genome, imaging and detailed blood profiles, all linked to medical follow-up. This layering of data, on the scale of half a million people, is what makes it unique.
| Data layer | Scope | Key date |
|---|---|---|
| Questionnaires, physical measurements, blood and urine samples | All 500,000 participants at enrollment | 2006–2010 |
| Genome | Genotyping, then whole-genome sequencing of all 500,000 participants | Data released in 2023, “the largest number of whole genomes ever released for medical research” |
| Imaging (MRI of the brain, heart and whole body, bone densitometry) | Target of 100,000 participants; repeat imaging planned for 10,000 of them two to three years later | More than 25,000 participants scanned by mid-2018 |
| Nuclear magnetic resonance (NMR) metabolomics, Nightingale Health | 249 measurements per sample; about 121,000 profiles released in 2021, then all 500,000 participants | Complete dataset made available on November 20, 2025 |
| Follow-up | Hospital records, registries, causes of death | At least 30 years |
The metabolomic layer deserves a closer look: it is the one found in some personalized health check-ups. The Nightingale Health NMR panel quantifies 249 measurements from a single blood sample: lipoproteins by class and size, fatty acids, amino acids, glycolysis metabolites, ketone bodies and an inflammation marker, GlycA. Nightingale states that it is the only “omics” technology measured on every sample in the cohort. We describe the technique in our article on NMR metabolomics.
What has UK Biobank revealed about prevention?
UK Biobank has led to three advances of direct use for prevention: an atlas linking each blood biomarker to hundreds of diseases, the demonstration that a metabolomic profile improves the 10-year prediction of several common diseases, and the genetic mapping of these biomarkers. This work is published in peer-reviewed journals and can be reproduced by other teams.
A biomarker–disease atlas
In 2023, Julkunen and colleagues published in Nature Communications an atlas of associations between the 249 NMR biomarkers and the prevalence, incidence and mortality of more than 700 common diseases, in 118,461 participants. It is the reference that makes it possible to say, for a given marker, which diseases it is associated with and in which direction.
Predicting several diseases from a single blood sample
In 2022, Buergel and colleagues (Nature Medicine) trained a model on 168 metabolic markers measured in 117,981 participants, with about 1.4 million person-years of follow-up, to estimate the 10-year risk of 24 common diseases: metabolic, vascular, respiratory, musculoskeletal and neurological diseases, and cancers. The “metabolomic profile” matched or outperformed established predictors for 15 of these diseases, and added predictive information to the full set of clinical variables for eight of them, including type 2 diabetes, dementia and heart failure. The results were validated in four independent cohorts. Conversely, no significant association was found for breast cancer: the authors themselves speak of the “potential and limitations” of the approach.
The genetics of biomarkers
In 2024, Karjalainen and colleagues (Nature) characterized the genetic determinants of 233 metabolic biomarkers in 136,016 people from 33 cohorts, identifying more than 400 genomic regions involved. This work helps distinguish, among the associations observed, those that reflect a causal mechanism.
Metabolomic age
Metabolomic age models, which estimate a person’s age from their blood profile and derive a gap with their chronological age, were also developed on UK Biobank. We discuss their significance and limits in our article on metabolic age.
What are the limits of UK Biobank?
UK Biobank has three well-documented limitations: its participants are healthier than the general population, they are overwhelmingly British and of European ancestry, and most measurements were taken only once, between the ages of 40 and 69. None of these invalidates the cohort; all of them shape how its results should be read.
The “healthy volunteer” effect
To build the cohort, 9.2 million people aged 40 to 69 were invited; 503,317 took part, a participation rate of 5.45% (Fry et al., American Journal of Epidemiology, 2017). Compared with the general population, participants were older, more often women, lived in less deprived areas, were less often obese, smokers or daily drinkers, and reported fewer diseases. Between the ages of 70 and 74, their all-cause mortality was 46.2% lower in men and 55.5% lower in women than in the general population, and their cancer incidence was 11.8% and 18.1% lower, respectively. This is what epidemiologists call healthy volunteer selection bias.
The practical consequence is twofold. On the one hand, the absolute risk levels observed in UK Biobank cannot be applied as they stand to a general population: they are too low. On the other hand, the associations between a risk factor and a disease remain usable: in 2020, Batty and colleagues (BMJ) compared UK Biobank with representative British surveys whose average response rate was 68% and found consistent associations between risk factors and mortality, concluding that “despite a very low response rate, risk factor associations in the UK Biobank seem to be generalisable.”
A British population, mostly of European ancestry
The volunteers were largely “healthy, wealthy and white, of European descent,” and the cohort is not representative of the diversity of the UK population. A model trained on these data may be less well calibrated in people of other ancestries, whose metabolic and genetic profiles differ in part. This is one of the reasons why the most robust publications validate their results in other cohorts.
A snapshot taken between the ages of 40 and 69
The cohort says nothing about adults under 40, and most biomarkers were measured once, at enrollment. The subsets with repeat measurements (imaging, a second blood sample) are valuable but partial. Finally, an observed association is not causation: the fact that GlycA is associated with future diseases does not mean that lowering GlycA prevents them.
What does UK Biobank change for a patient in France?
For a patient in France, UK Biobank changes the available knowledge, not the course of action: the biomarkers that matter are better identified and risk models are better supported, but any estimate derived from a population of British volunteers must be interpreted by a doctor, within the French care pathway.
Better-supported tools
The metabolomic panels offered in France use the same NMR platform and the same 249 measurements as the cohort: an individual result can therefore be read against a solid reference. We describe these markers in detail in the guide to blood biomarkers. Likewise, the European SCORE2 and SCORE2-OP scores (2021) are calibrated by region, with France in the low-risk region, as we explain in the article on 10-year cardiovascular risk.
An interpretation that remains medical
A probability calculated from a cohort applies to a group of similar people, not to you in particular. Healthy volunteer bias, population differences and the lack of French validation for some models call for caution: an “at-risk” profile is not a diagnosis, and we do not write that a test “detects” a disease. What UK Biobank studies show is that a metabolomic profile improves the prediction of 10-year risk for several diseases, at the population level. What this information means for you is a matter of predictive medicine as practiced by a doctor: weighing it against your medical history, your clinical measurements and your priorities.
The covered pathway first
In France, the entry point remains the prevention check-up offered at four key ages, fully covered (100%) by French National Health Insurance (Assurance Maladie), with no upfront payment, which leads to a Personalized Prevention Plan sent to your regular doctor (médecin traitant) unless you object. A more in-depth analysis can then be discussed, if it could change a decision. This is the principle of quaternary prevention, which we develop in our personalized health guide.
Key takeaways
- UK Biobank follows about 500,000 Britons aged 40 to 69, recruited between 2006 and 2010, for at least thirty years, with whole genomes, imaging and 249 metabolomic biomarkers now available for the entire cohort.
- It has produced a biomarker–disease atlas (more than 700 diseases), shown that a metabolomic profile improves the 10-year prediction of several common diseases, and mapped the genetics of these biomarkers.
- Its participants are healthier than the general population (5.45% participation rate, markedly lower mortality): the associations are generalizable, but absolute risk levels are not without recalibration.
- For a patient in France, these results inform the interpretation of a check-up; they replace neither the doctor nor the covered prevention pathway.
What Sokrate lets you do
The Sokrate service prepares your prevention check-up online through an adaptive questionnaire, after which a doctor writes and signs your Personalized Prevention Plan, sent to your regular doctor unless you object. The check-up carried out by an authorized professional is fully covered (100%) by French National Health Insurance, with no upfront payment. As an option, and in agreement with your doctor, an NMR metabolomic analysis of 249 biomarkers, on the platform used in UK Biobank and interpreted by a doctor, can refine the estimate of your risks. Join the waitlist or find out how the pathway works.