Background: Several epidemiological studies have shown an association between increased fasting and 2-hour plasma glucose (FPG and 2hPG), hemoglobin A1C (A1C), and fasting insulin (FI) and risk of atrial fibrillation (AF). We aimed to assess whether this relation is causal using genetic data from large populations of individuals of European descent.
Methods: Two-sample Mendelian randomization (MR) was used to obtain estimates of influence of the risk factors (FPG, 2hPG, A1C and FI) on risk of the outcome (AF). Instrumental variables (IVs) were constructed using the previously reported single nucleotide polymorphisms (SNPs) associated with each risk factor at p-value<e-8 from the Meta-Analyses of Glucose and Insulin-related traits Consortium (MAGIC) with up to 133,000 nondiabetic participants. Selected SNPs were checked for pleiotropy using the UK Biobank dataset and those associated with the confounders at p-value<e-8 were excluded. AF consortium (AFGen) summary statistics with 22,346 AF cases and 132,086 controls was used for the outcome data analysis.
Results: Two-step MR with inverse variance weighted methods did not show any causal association between diabetes related phneotypes and AF [FPG (30 SNPs, p=0.7), 2hPG (7 SNPs, p=0.3), A1C (11 SNPs, p=0.3), or FI (13 SNPs, p=0.9)]. None of the IVs showed horizontal pleiotropy indicating that the used genetic variants affected the outcome only via their effect on the diabetes related phenotypes (p-values of 0.56, 0.09, 0.099, and 0.31 respectively for FPG, 2hPG, A1C, and FI). Calculated power based on the original set of SNPs in the MAGIC and estimated odds ratio of 1.2 were 99, 94, 99, and 84% respectively for FBG, 2hPG, A1C, and FI.
Conclusion: This MR analysis does not support a causal role between genetically programmed diabetes related traits (FPG, 2hPG, A1C, or FI) and AF. These data suggest that drug treatment to reduce dysglycemia in nondiabetics may not be an effective strategy for AF prevention.
H. Harati: None. D. Zanetti: None. E. Ingelsson: None. J. Knowles: None.