Brown fat is difficult to study at population scale because it does not announce itself in a routine blood test. It appears as metabolically active tissue on certain PET–CT scans, often around the neck, collarbone and armpit. A new study used a large archive of clinical scans and machine-assisted measurement to turn those scattered signals into a cohort of 101,243 patients.[1][2]
Across diagnoses in the cancer-centre cohort, cardiovascular and psychiatric diseases were less prevalent among people whose scans showed brown adipose tissue. The researchers then semi-automatically quantified brown-fat activity in roughly 8,700 people. Higher activity, especially in the axillary depot, was associated with lower odds of cardiometabolic disease and with more favourable metabolic, inflammatory and cardiac measures.[1]
The study climbs an evidence ladder
Four steps from scans to a drug hypothesis
| Step | Evidence | What it can support |
|---|---|---|
| Large retrospective cohort | 101,243 patients with clinically acquired PET–CT scans | Associations between visible brown fat and recorded diagnoses or measurementsSource 1 |
| Quantitative phenotyping | Machine-assisted activity measurement in about 8,700 patients | Depot-specific and activity-level comparisons rather than presence aloneSource 1 |
| Medication analysis | Associations between recorded drugs and brown-fat presence | Candidates for mechanistic follow-up; not proof that a medicine caused the tissue changeSource 1 |
| Cells and mice | Losartan experiments on brown adipocytes and animal metabolism | Biological plausibility and proof of concept outside a human treatment trialSource 1Source 2 |
The machine-learning component was not asked to diagnose a disease. It helped measure the amount and activity of brown fat across many images, reducing the manual burden of delineating tissue depot by depot. About 8.6% of the full cohort entered this quantified subset, so results based on activity are drawn from a smaller group than the presence-or-absence comparisons.[1]
The medication signal came from the records
The analysis found positive associations between brown-fat presence and several medicines. The team focused on angiotensin receptor blockers because their biological targets or pathways were plausible in adipocytes. Losartan use was associated with brown-fat-positive status with an odds ratio of 1.36, while valsartan’s odds ratio was 1.52. Both confidence intervals excluded one in the reported analysis.[1]
The human medication associations
| Measure | Value | Context |
|---|---|---|
| Losartan | OR 1.36 | 95% CI 1.13–1.63; P = 0.006Source 1 |
| Valsartan | OR 1.52 | 95% CI 1.14–2.02; P = 0.02Source 1 |
| Irbesartan | Positive trend | Not statistically significant in the reported analysisSource 1 |
The laboratory work tests plausibility, not human benefit
In cultured brown adipocytes, higher concentrations of losartan increased expression of UCP1, a protein central to heat production, and raised oxygen consumption through a pathway involving PPARγ. In mice, losartan increased energy expenditure under the tested conditions. Those experiments make the record-based association more biologically coherent.[1][2]
They do not establish that a person should take losartan to activate brown fat. The human data are retrospective, the cohort comes from a cancer centre, scans were acquired for clinical purposes without a standard cold-stimulation protocol, and medication use can be linked to many characteristics that are difficult to remove completely. The next decisive evidence would be a prospective human study designed around brown-fat activation and clinically meaningful outcomes.[1]
Why the workflow is worth studying
The study is a good example of reverse translation. Instead of beginning with one molecule in a mouse and hoping it matters to people, the researchers began with a large human clinical archive, used AI to expose a phenotype, identified a medication signal and then tested a mechanism in cells and animals. The chain does not remove uncertainty; it makes each uncertainty easier to locate.[1][2]
Sources & further reading
- Artificial intelligence-enabled clinical phenotyping of human brown fat Nature · 14 Oct 2026 · Accessed 16 Oct 2026
- A blood-pressure drug may activate health-promoting brown fat National Institutes of Health · 14 Oct 2026 · Accessed 16 Oct 2026