Genetically predicted adipose tissue distribution influences the risk of atherosclerosis.

Abstract

Truncal obesity is a key modifiable risk factor in cardiovascular disease (CVD). It is a highly complex phenotype, and numerous factors such as the distribution of adipose tissue may influence risk of CVD.1 However, while some studies have suggested that gluteofemoral fat deposits may serve a protective role, others have offered conflicting results.2,3 As observational studies may be vulnerable to confounding from environmental factors, we conducted a Mendelian randomization study investigating the effect of regional adipose tissue distribution on the risk of different types of atherosclerosis and heart failure (HF). The Mendelian randomization approach leverages genetic variation to infer causal associations as genetic variants in the human genome are inherited randomly, independently of environmental factors.4 All data used in this study were publicly available. First, we obtained summary statistics from a recent genome-wide association study on visceral adipose tissue (VAT), abdominal subcutaneous adipose tissue (AAT), and gluteofemoral adipose tissue (GFAT).5 Adipose tissue traits were determined by magnetic resonance imaging (MRI) of up to 39 076 individuals from the UK Biobank and were adjusted for age, sex, BMI, height, and 10 principal components. Instrument variables were selected from single-nucleotide polymorphisms (SNPs) that were robustly associated with the traits on a genome-wide significant level (P < 5 × 10−8). Instrument variables were pruned for linkage disequilibrium (LD), using a 10 megabase (Mb) window. For genetically correlated variants (R2 > 0.001), the variant with the smallest P-value was selected. We considered myocardial infarction (MI), large artery atherosclerotic stroke (LAS), and peripheral artery disease (PAD) as proxies for atherosclerosis. We acquired MI and LAS summary statistics from the CARDIoGRAMplusC4D consortium6 and the European subset of the GIGASTROKE consortium.7 Summary statistics for PAD and HF were obtained from the FinnGen study.8 There was no sample overlap between exposure and outcome datasets. For instrument variables not present in the outcome datasets, we identified suitable proxy SNPs in LD (r2 > 0.8).