Generalized Summary Data-based Mendelian Randomization Method And Its Application In Bidirectional Causal Association Between Type 2 Diabetes And Five Psychiatric Disorders

Objectives:In order to evaluate the performance of the four Mendelian randomization methods of GSMR,IVW,MR-Egger and MBE under different parameter settings through simulation study,and to explore and provide suggestions for selecting appropriate Mendelian randomization methods for causal inference.To assess the causal effects between type 2 diabetes(T2D)and five psychiatric disorders(anxiety disorders,bipolar disorder,major depression,schizophrenia,and obsessive-compulsive disorder),as well as for providing genetic support for the association between T2 D and five psychiatric disorders.Methods:According to the basic principles of GSMR,IVW,MR-Egger and MBE,we seted up six simulation situations,which were used to test sample size,pleiotropy,linkage disequilibrium and InSIDE hypothesis for the influence of each method,respectively.And mean causal effect value,standard deviation,standard error,mean square error,type ? error rate and power were used as indicators to evaluate the advantages and disadvantages of GSMR,IVW,MR-Egger and MBE under different parameter settings.The primary analysis was performed using the GSMR method,while the other three methods were used as sensitivity analyses to estimate the causal association between T2 D and five psychiatric disorders.Results:Based on the results of the simulation study,we found that the GSMR outperformed other three methods in terms of type ? error rate in most situations,for the MBE presented too conservative type ? error rate.Meanwhile,the GSMR performed best and MR-Egger performed worst in terms of the efficiency of finding causal effects.Under the condition that all the instrument variables were valid,all the indexes of the GSMR were similar with the IVW,and the requirements for sample size of the GSMR and the IVW were lower than that of the MR-Egger and MBE.The GSMR method shows good robustness in the presence of pleiotropy and linkage disequilibrium between instument variables,and its standard deviation,standard error and standard mean square error outperformed the other three methods.Futhermore,IVW,MR-Egger,MBE and GSMR methods all have different degrees of dependence on the InSIDE hypothesis,and whether the InSIDE hypothesis is satisfied or not has a significant impact on IVW and MR-Egger.Besides,we also found a causal relationship that T2 D has causal effect on anxiety disorders(OR: 1.153,95%CI: 1.057-1.250,p=0.004),however,no evidence of a reverse causality for T2 D with anxiety disorders was found(p=0.231).Meanwhile,no evidence was found for bidirectional causal association between T2 D and bipolar disorder,major depression,schizophrenia,and obsessive-compulsive disorder in the study.Conclusions:Comparing to other three MR methods,GSMR showed higher efficiency and robustness,so the GSMR method should be more widely used in the Two-Samples MR analysis to infer the risk factor whether play a causal role for the development of the disease.In the present study,we utilized the GSMR method for causal inference and found that genetic susceptibility of T2 D is associated with an increased risk of anxiety disorders.Combined with the findings of the previous observational studies,this study provides evidence for the causal relationship between T2 D and anxiety disorders risk.