Associations of visceral adipose tissue with bone mineral density and fracture: observational and Mendelian randomization studies

Abstract

Background: The associations between visceral adipose tissue (VAT) and bone mineral density (BMD) or fracture have been controversial and the causality of the associations remains to be assessed. This study aimed to explore the associations of VAT^ (predicted value of VAT mass) with BMD and fracture risk in men and women, and to examine their potential causation by two-sample Mendelian randomization (MR) analyses. Methods: UK Biobank is a large, population-based prospective cohort study that recruited more than 500,000 participants aged 40–69 in the United Kingdom from 2006 to 2010. In this study, we used a validated and reliable prediction model to estimate the VAT amount of the participants. On this basis, linear and nonlinear multivariable statistical models were used to explore the association of VAT^ with BMD and fracture risk in different groups of sex and BMI. In observational analyses, the multivariable linear regression model and Cox proportional-hazards model were used to assess VAT^ association with BMD and fracture risk, respectively. Inverse variance weighting was used as the main result of MR analysis. Results: In 190,836 men, an inverted U-shaped association was observed between VAT^ and heel BMD (P for nonlinearity < 0.001), with a turning point of VAT^ = 1.25 kg. Per kg increase in VAT^ was associated with a 0.13 standard deviation (SD) increase in heel BMD (P = 1.5 × 10−16) among men with lower amounts of VAT^, and associated with a 0.05 SD decrease in heel BMD (P = 1.3 × 10−15) among men with higher amounts of VAT^. In 193,592 women, per kg increase in VAT^ was monotonically associated with a 0.16 SD increase in heel BMD (P = 1.2 × 10−136, P for VAT^-sex interaction = 8.4 × 10−51). During a median follow-up of 8.2 years, VAT^ was associated with lower risks of hip fractures in the overall men and women (P for VAT^-sex interaction = 1.9 × 10−4 for total fractures; 1.5 × 10−4 for other fractures). There were significant interactions of VAT^ and BMI on heel BMD and fracture risks in men only (P for VAT^-BMI interaction = 5.9 × 10−31 for heel BMD; 2.7 × 10−4 for total fractures; 5.7 × 10−3 for hip fractures; 6.8 × 10−3 for other fractures). In two-sample MR analyses, evidence of causality was not observed between VAT^ and DXA-derived BMD or fractures. Conclusions: These novel findings demonstrated gender-dependent associations of VAT^ with BMD and fracture risk, with the association in men being modified by adiposity. Evidence of causality was not observed, suggesting that the observational association of VAT^ with BMD and fracture risk could be the result of confounding.