1408-P: Biobank Scale Identification of Exposome Risk Factors in Cardiometabolic Disease and Their Mediators

Cardiovascular disease and diabetes are influenced by potentially many exposures; however, traditional epidemiological studies examine but a few candidate factors at a time. Here, we demonstrate a pipeline for identifying exposures, made efficient by integrating genetic data at biobank scale. First, we conducted an Exposure-Wide Association Study (ExWAS) to systematically identify observational associations across 362 exposures in type 2 diabetes [T2D] and coronary artery disease [CAD] in participants of the UK Biobank. Causality estimates between exposure-disease phenotype pairs using bi-directional Two-Sample Mendelian Randomization [MR] across two large biobanks, UK Biobank and FinnGen, were computed. For CAD, we identified 172 (47.5%) FDR-significant exposure factors and for type 2 diabetes we identified 224 (61.9%) factors, finding 157 in both. We were able to deduce genetic-based causality between 14 (out of 172 ExWAS-identified) in CAD and 16 (out of 224) in T2D. Among the MR-validated associations for CAD, we report the interquartile range (IQR) of ExWAS and MR-derived odds ratios to be [0.742, 0.892] and [0.389, 0.711], respectively. For T2D, we report the IQR to be [0.658, 0.897] and [0.277, 0.669] for ExWAS and MR-derived odds ratios (OR), respectively. For example, we found having a college or university degree is causally inversely associated with CAD (MR beta estimate = -0.569, MR p-value = 3.51x10-5). However, we found low specificity (37.3 % for CAD and 26.3% for T2D) suggesting that most ExWAS-identified (or observational) associations are confounded. We find three MR-validated associations for T2D and one MR-validated association for CAD to be strongly mediated by causal risk factors (e.g., Body Mass Index[BMI]). Biobank samples, replete with clinical, exposure and genetic information enhance identification of exposures for metabolic disease and their mediators.