Enabling precision cardiometabolic disease prevention through artificial intelligence

• Metabolic Diseases
• Digital Health

Researchers from the LIH Department of Population Health (DoPH) and from the LIH Quantitative biology Unit (QBU) leveraged advanced artificial intelligence and machine-learning approaches to classify a sample of individuals representative of the Luxembourgish population into distinct risk groups, paving the way towards more targeted strategies for the prevention of cardiometabolic diseases at a population level. The findings were published on August 6th 2021 in the Nature Portfolio journal “Scientific Reports”.    

The rapid increase in the incidence of cardiometabolic conditions, such as type 2 diabetes and hypertension, urgently calls for better prevention strategies, moving from a one-size-fits-all to a precision approach in the general population. This should take into account the high variability observed in individuals in terms of genetic profiles, inflammation, oxidative stress, insulin resistance and sugar levels, and ensuing risk of developing cardiometabolic disorders. 

“In this context, artificial intelligence and machine-learning approaches can prove to be very valuable to identify subgroups of the population with different cardiometabolic risk profiles”, explains Dr Guy Fagherazzi, Director of DoPH and first author of the publication.

To this end, the research team leveraged a technique known as semi-supervised cluster analysis in the general population and at a large scale. Using the unique set of 29 different cardiometabolic characteristics  available from the nationwide population-based ORISCAV-LUX 2 study, they set about classifying the Luxembourgish population in terms of cardiometabolic profiles, driven mainly by body mass index (BMI) − the most frequently used indicator of adiposity and an established risk factor in numerous cardiometabolic disorders − and glycated hemoglobin (HbA1c), a reliable indicator of blood sugar levels also correlated with many cardiometabolic conditions. 

The scientists observed that the 1356 participants considered could be grouped into four distinct clusters. Specifically, 729 individuals (i.e.  53.8% of the study population) belonged to Cluster 1, generally characterised by a young age, low blood sugar levels, low BMI, low adiposity, healthy cardiovascular parameters and better lifestyle indicators. “Individuals in this ‘cardiovascularly healthy’ cluster therefore reported the lowest cardiovascular age and a 0% average 10-year cardiovascular risk”, says Dr Fagherazzi. The 508 participants in Cluster 2 “Family history – Overweight – High Cholesterol” were mainly overweight, with low HbA1c levels, elevated total and low-density lipoprotein (LDL) cholesterol levels and a high frequency of family history of diabetes and high blood pressure. “The average 10-year cardiovascular risk for Cluster 2 was therefore higher than Cluster 1, at 2%”, adds Dr Fagherazzi. 91 participants belonged to Cluster 3 “Severe Obesity – Prediabetes – Inflammation”, characterised by obesity and even higher BMI and HbA1c levels, as well as the most elevated level of inflammation, while the 28 Cluster 4 “Diabetes – Hypertension – Poor Cardiovascular Health” members were mainly overweight or obese, displaying high blood sugar and fat levels and suffering from diabetes and hypertension. They therefore reported the highest cardiovascular age, with the average 10-year cardiovascular risk reaching 15%.  

Our work provides an in-depth characterization and thus, a better understanding of cardiometabolic health in the general population. Our data suggest that such a clustering approach could now be used to define more targeted and tailored strategies for the personalised prevention of cardiometabolic diseases at a population level”

states Dr Fagherazzi.

The results could therefore indicate that Cluster 1 individuals could benefit from a general prevention strategy, while the other 3 clusters may benefit from a more personalised and intensive approach to improve their health, focusing for instance specifically on overweight/obesity management, personalised cholesterol treatment, a targeted lifestyle management strategy, or even bariatric surgery, according to the particular cardiovascular health parameters and socioeconomic profiles that characterise them.

If externally validated, general practitioners could one day rely on this profiling to have a better picture of a new patient and to optimise several cardiometabolic parameters simultaneously

concludes Dr Fagherazzi.

In future studies, wearable devices could also be used to collect objective measures of additional factors of interest, such as physical activity, mental health and sleep quality, which may be valuable information to complement the cluster analysis.