Lipoprotein(a) is not a causal factor for the development of diabetes - a two-sample Mendelian randomization study

Abstract

Abstract Introduction The association between lipoprotein(a) [Lp(a)] and coronary artery disease is well known. Nevertheless, because physiological function of Lp(a) is not well understood, it is not clear what sort of adverse effects therapies aimed at Lp(a) reducing might have. One of the associations that was observed for the reduced Lp(a) levels was the increased incidence of diabetes. As genetic factors have a clear impact on Lp(a) levels, we applied Mendelian Randomization [MR] approach as a tool to determine the impact of Lp(a) on the risk of diabetes onset. Purpose To assess the impact of Lp(a) levels on the incidence of diabetes. Methods We conducted a two-sample MR study. Publicly available datasets were used as sources for summary statistics of genetic correlations. UK Biobank was the source of correlations for Lp(a) levels and FinnGen was used as the source of genetic correlations for type 2 diabetes. Only variants in the LPA gene region were utilized. Inverse-variance-weighted mean method was used for primary analysis. Secondary analyses were conducted by weighted median, mode, and MR-Egger methods. Sensitivity analyses were performed by leaving out random 30% of the variants as well as by using LPA gene variants that were previously published as significantly affecting Lp(a) levels. Results Data regarding genetic determinants of Lp(a) levels were available for 318922 participants, while summary statistics for type 2 diabetes were available for 49114 cases and 283207 controls. Primary analysis indicated that Lp(a) levels are not correlated with diabetes [p=0.362]. Genetic associations of included variants with Lp(a) levels (horizontal axis) and with type 2 diabetes mellitus (vertical axis) are presented in Figure. Secondary analyses were consistent with this finding [median p=0.100; mode p=0.156; MR-Egger p=0.243]. Selected variants did not show appreciable pleiotropy as indicated by MR-Egger intercept test [p=0.450]. Sensitivity analyses did not result in statistically significant findings [leave-30%-out analysis IVM p-values: 0.153-0.960]. Conclusions Lp(a) levels determined by genetic background are not correlated with type 2 diabetes risks. This indicates that Lp(a) levels has likely no causal role in type 2 diabetes onset. This suggests that treatments aimed at reducing Lp(a) levels will not likely increase the risk of developing type 2 diabetes.