Causal associations of insulin and Lp(a) levels: a Mendelian randomization study

Abstract Background Numerous observational studies have demonstrated circulating lipoprotein(a) [Lp(a)] to be inversely related to occurrence of type 2 diabetes [T2D]. However, this is not consistently supported by Mendelian randomization [MR] studies. In vitro studies have shown insulin to downregulate Lp(a) expression, which might be another mechanism behind the correlation observed between Lp(a) and T2D. Purpose In this MR study, we investigated the influence of genetically predicted levels of insulin on genetically predicted Lp(a) levels to elucidate the potential causal links between Lp(a) and diabetes. Methods Independent genetic variants associated with insulin levels were acquired from a meta-analysis of genome-wide association studies (GWAS, N=151,013). Summary data for Lp(a) were acquired from a GWAS study in the UK Biobank (N=361,194). Inverse-variance-weighted (IVM) method was used to perform a two sample MR study. Sensitivity analysis was conducted via MR Egger, weighted median (WME), as well as leave-one-out analysis. Results Genetically predicted insulin levels were negatively associated with Lp(a) levels according to IVM estimate (p=0.003). In sensitivity analysis, WME supported this result (p=0.0002), while MR-Egger was not statistically significant (p=0.11). Leave-one-out analysis did not show the results to depend on any single variant. Conclusion This MR study provides robust evidence supporting the association between plasma insulin levels and decreased Lp(a) concentration. These findings suggest that hyperinsulinemia triggered by insulin resistance can be in part responsible for the observed negative correlation between low Lp(a) and T2D risk. Figure 1. Genetic associations between insulin levels and Lp(a). Each genetic variant included in the analysis is represented as a point + 95% CI. Localization on the horizontal axis represents the correlation of the variant with exposure (plasma fasting insulin, inverse variance normal transformed values). Localization on the vertical axis represents the correlation of the variant with outcome (Lp(a), natural log transformed values). Lines represent estimates of different MR methods.Figure 1.