Brown fat: AI analyses scans from 101,000 patients and links it to lower rates of diabetes, heart disease and depression

Published in *Nature*, the study uses AI to quantify brown adipose tissue across more than 101,000 PET scans and identifies a potential drug candidate, losartan, which has yet to be tested in humans.

A team led by Gómez-Banoy used 101,243 18F-FDG PET-SCAN scans performed between June 2009 and July 2024 to map brown fat, the tissue that burns energy to produce heat. An AI tool was used to identify and quantify its metabolic activity in six distinct anatomical areas. A total of 8,709 patients had detectable brown fat.

Marked associations

In these people, the chances of having type 2 diabetes were 58% lower, and those of chronic ischemic heart disease were 47% lower. The researchers also noted a better cholesterol profile, fewer inflammatory markers and a lower prevalence of depressive episodes and anxiety disorders. In the laboratory and in mice, a common antihypertensive drug, losartan, stimulated the thermogenic activity of this tissue, making it an avenue for research.

The authors remain cautious: the study is retrospective, conducted on the population of a cancer center, without exposure to cold before the examinations, and glucose uptake is only an indirect indicator of thermogenesis. It establishes correlations, not a cause-and-effect link, and the interest of losartan in humans remains to be demonstrated by clinical trials.