If you’re obese, you’re at a higher risk for cardiovascular disease, Type 2 diabetes and other complications. But some people, despite having a high body mass index (BMI), have a low risk for these complications. On the other hand, about one in five people with a normal BMI are at high risk – for instance, a person might have a much higher amount of fat or sugar in their blood than you would expect when considering their body weight alone, which could put them at a higher-than-expected risk for heart disease or diabetes.
The IMI Sophia project used machine learning techniques to analyse health data from a database of 170 000 adults from the UK, the Netherlands and Germany. They also developed powerful algorithms to cluster people whose risk profiles didn’t match what was expected given their body weight.
The study defined five subgroups of people whose risk for obesity complications and BMI are not aligned. For instance, about 5% of women and 7% of men had a profile with high “bad” cholesterol (LDL), high levels of fat in the blood, and had higher blood pressure than expected for their weight. This indicates that, despite their weight being within the normal range, these people are at risk of a cardiovascular event, diabetes, or another complication commonly linked to obesity.
This finding can help to identify people who are at risk of developing cardiovascular disease, Type 2 diabetes or other complications commonly associated with obesity, but who are not obese. Since a lot of these complications are preventable, it is possible to reduce the number of people who go on to develop these complications.
What’s more, better and more precise treatments can now be developed that are targeted for a specific subtype.
“The subgroups of people with obesity identified by the IMI SOPHIA project allows us to understand why two people with exactly the same body mass index and the same body fat distribution can have completely different obesity related complications,” said Carol Le Roux, project coordinator of SOPHIA.
“We are now starting to no longer think of obesity as one disease but rather as multiple different diseases that can lead to the same level of adiposity. This will improve how patients, clinicians, payers, and regulators view obesity as a subset of disease.”
Ali Farzaneh, one of the lead researchers involved in the study, said that these results illustrate how powerful machine learning techniques and personalised medicine approaches can be for health research.
“These findings redefine the understanding of obesity by identifying distinct subtypes that drive cardiometabolic risk, offering new precision tools for early disease prevention and intervention. This work demonstrates the potential of advanced algorithms and large-scale data to improve health outcomes.”
SOPHIA is supported by the Innovative Medicines Initiative, a partnership between the European Union and the European pharmaceutical industry.