Obesity and cardiometabolic disease often, but not always, coincide. Distinguishing subpopulations within which cardiometabolic risk diverges from the risk expected for a given body mass index (BMI) may facilitate precision prevention of cardiometabolic diseases. Accordingly, we performed unsupervised clustering in four European population-based cohorts (N ≈ 173,000). We detected five discordant profiles consisting of individuals with cardiometabolic biomarkers higher or lower than expected given their BMI, which generally increases disease risk, in total representing ~20% of the total population. Persons with discordant profiles differed from concordant individuals in prevalence and future risk of major adverse cardiovascular events (MACE) and type 2 diabetes. Subtle BMI-discordances in biomarkers affected disease risk. For instance, a 10% higher probability of having a discordant lipid profile was associated with a 5% higher risk of MACE (hazard ratio in women 1.05, 95% confidence interval 1.03, 1.06, P = 4.19 × 10⁻¹⁰; hazard ratio in men 1.05, 95% confidence interval 1.04, 1.06, P = 9.33 × 10⁻¹⁴). Multivariate prediction models for MACE and type 2 diabetes performed better when incorporating discordant profile information (likelihood ratio test P < 0.001). This enhancement represents an additional net benefit of 4−15 additional correct interventions and 37−135 additional unnecessary interventions correctly avoided for every 10,000 individuals tested.

Obesity is considered by many as a lifestyle choice rather than a chronic progressive disease. The Innovative Medicines Initiative (IMI) SOPHIA (Stratification of Obesity Phenotypes to Optimize Future Obesity Therapy) project is part of a momentum shift aiming to provide better tools for the stratification of people with obesity according to disease risk and treatment response. One of the challenges to achieving these goals is that many clinical cohorts are siloed, limiting the potential of combined data for biomarker discovery. In SOPHIA, we have addressed this challenge by setting up a federated database building on open-source DataSHIELD technology. The database currently federates 16 cohorts that are accessible via a central gateway. The database is multi-modal, including research studies, clinical trials, and routine health data, and is accessed using the R statistical programming environment where statistical and machine learning analyses can be performed at a distance without any disclosure of patient-level data. We demonstrate the use of the database by providing a proof-of-concept analysis, performing a federated linear model of BMI and systolic blood pressure, pooling all data from 16 studies virtually without any analyst seeing individual patient-level data. This analysis provided similar point estimates compared to a meta-analysis of the 16 individual studies. Our approach provides a benchmark for reproducible, safe federated analyses across multiple study types provided by multiple stakeholders.

Obesity and cardiometabolic disease often, but not always, coincide. Distinguishing sub-populations within which cardiometabolic risk diverges from the risk expected for a given body mass index (BMI) may facilitate precision prevention of cardiometabolic diseases. Accordingly, we performed unsupervised clustering in four European population-based cohorts (N ~ 173K). We detected five discordant profiles consisting of individuals with cardiometabolic biomarkers higher or lower than expected given their BMI, in total representing ~20% of the total population. Persons with discordant profiles differed from concordant individuals in prevalence and future risk of major adverse cardiovascular events (MACE) and diabetes. Subtle BMI-discordances in biomarkers affected disease risk. For instance, a 10% higher probability of having a discordant lipid profile was associated with a 5% higher risk of MACE (HR in women: 1.05, 95% CI: 1.03, 1.06, p = 4.19x10-10; HR in men: 1.05, 95% CI: 1.04, 1.06, p= 9.33x10-14). Multivariate prediction models for MACE and diabetes performed better when incorporating discordant profile information (likelihood ratio test P < 0.001). This enhancement represents an additional net benefit of 4 to 15 additional correct interventions, and 37 to 135 additional unnecessary interventions correctly avoided for every 10,000 individuals tested.