1907-P: Using Human Genetics to Test the Disease Consequences of Varying Testosterone Levels in Men and Women

Circulating testosterone and SHBG levels are associated with type 2 diabetes (T2D). These associations differ between sexes and by total and bioavailable testosterone. For example, in UK Biobank a 1-SD raised bioavailable testosterone is associated with lower risk of T2D in men (OR=0.90 [0.88,0.92]) but higher risk in women (OR=1.24 [1.20,1.29]). Genetic studies provide more robust tests of causality and have indicated that higher circulating SHBG is protective of T2D (OR=0.92 [0.88,0.96], Perry et al 2010), but were limited to few genetic variants. In this study we aimed to: i) expand the set of genetic variants associated with circulating testosterone; ii) separate variants into clusters representing SHBG or direct testosterone effects; and iii) perform Mendelian Randomisation studies to better understand the potential causal effects of SHBG and testosterone on T2D and related traits. We used genome-wide analyses of sex hormones from 425,097 UK Biobank participants. We identified 2,571 genome-wide significant (P<5x10^-8) associations (vs. 17 pre-2019). The genetics of testosterone levels were very different between men and women (genetic correlation ∼0). To identify variants with specific testosterone-increasing effects independent of SHBG, we used a hierarchical-clustering strategy. In men, 122 loci were associated with higher testosterone in a dose-response manner with no effect on SHBG, with 241 such signals in women. We found that SHBG and testosterone levels can influence T2D and related traits. A genetically determined 1-SD higher testosterone increased the risk of T2D (OR=1.37 [1.22,1.53]) and polycystic ovary syndrome (OR=1.51 [1.33,1.72]) in women, but reduced T2D risk in men (OR=0.86 [0.76,0.98]). Using clustered variants suggested that diabetes effects were primarily mediated by testosterone in men. Our findings provide insights into the disease impacts of testosterone and highlight the importance of sex-specific analyses of diabetes.