Identification of Putative Causal Relationships Between Type 2 Diabetes and Blood-Based Biomarkers in East Asians by Mendelian Randomization

Abstract

Observational studies revealed phenotypic associations between type 2 diabetes (T2D) and many biomarkers. However, causality between these conditions in East Asians is unclear. We leveraged genome-wide association study (GWAS) summary statistics of T2D (Ncase = 77,418; Ncontrol = 356,122) from the Asian Genetic Epidemiology Network (sample recruited during 2001-2011), and the GWAS summary statistics of 42 biomarkers (N = 12,303 to 143,658) from BioBank Japan (sample recruited during 2003-2008) to investigate causal relationships between T2D and biomarkers. Applications of Mendelian randomization (MR) approaches consistently revealed the genetically instrumented T2D associated with increased blood potassium (liability-scale β = [0.04, 0.10], p-value = [6.41×10-17, 9.85×10-5]) and decreased blood chloride (liability-scale β = [-0.16, -0.06], p-value = [5.22×10-27, 3.14×10-5]) whereas these two biomarkers showed no causal effects on T2D. Heritability estimation using summary statistics (ρ-HESS) and summary data-based Mendelian randomization (SMR) highlighted 27 genomic regions and three genes (MGAT1, TLE1, and HMGCR) interactively associated with the shared genetics underlying T2D and the two biomarkers. Thus, T2D may causally affect potassium and chloride in blood for East Asians. In contrast, the relationships from the potassium or chloride to T2D are not causal, suggesting the importance of monitoring the electrolyte disorder for T2D patients.

Inverse association between type 2 diabetes and hepatocellular carcinoma in East Asian populations

“…( Supplementary Table 1 ) Association analyses were adjusted for gender and body mass index(BMI). These instruments have also been used in previous MR studies ( 25 , 26 ). Additionally,We also used this summary-level data for T2D in the reverse MR analysis.…”

Genetic evidence for T-wave area from 12-lead electrocardiograms to monitor cardiovascular diseases in patients taking diabetes medications

“…The putative causal relationship between two traits was considered to be determined if consistent results were identified by more than four MR models with FDR < 0.05. Depending on the data type of the outcome and exposure trait, the MR estimates (i.e., β xy ) were converted to liability scale using one of the methods described by Byrne et al(29, 30). The related formulae of the methods were described in the Supplementary Material.…”

“…Depending on the data type of the outcome and exposure trait, the MR estimates (i.e., ߚ ௫௬ ) were converted to liability scale using one of the methods described by Byrne et al (29,30).…”