Multivariable Mendelian Randomization: The Use of Pleiotropic Genetic Variants to Estimate Causal Effects

Abstract

A conventional Mendelian randomization analysis assesses the causal effect of a risk factor on an outcome by using genetic variants that are solely associated with the risk factor of interest as instrumental variables. However, in some cases, such as the case of triglyceride level as a risk factor for cardiovascular disease, it may be difficult to find a relevant genetic variant that is not also associated with related risk factors, such as other lipid fractions. Such a variant is known as pleiotropic. In this paper, we propose an extension of Mendelian randomization that uses multiple genetic variants associated with several measured risk factors to simultaneously estimate the causal effect of each of the risk factors on the outcome. This “multivariable Mendelian randomization” approach is similar to the simultaneous assessment of several treatments in a factorial randomized trial. In this paper, methods for estimating the causal effects are presented and compared using real and simulated data, and the assumptions necessary for a valid multivariable Mendelian randomization analysis are discussed. Subject to these assumptions, we demonstrate that triglyceride-related pathways have a causal effect on the risk of coronary heart disease independent of the effects of low-density lipoprotein cholesterol and high-density lipoprotein cholesterol.

Genetic support for the causal role of insulin in coronary heart disease

“…1), which does not violate the causal inference. In this study, rather than excluding SNPs to decrease the risk for pleiotropy, we applied methods that take pleiotropic effects into account at the modelling phase [5,13,14].…”

“…To evaluate the causal effects of risk factor traits on CHD, we applied a statistical method that controls for pleiotropic effects, noted as 'weighted regression-based method' [14]. This method has been described in Do et al and Burgess et al [5,14]; in short, we extracted the SNP effect estimates for LDL-and HDL-cholesterol, triacylglycerols, BMI, WHR, fasting insulin and glucose, systolic and diastolic BP, and CHD from the summary result files of reference GWASs.…”

“…This method has been described in Do et al and Burgess et al [5,14]; in short, we extracted the SNP effect estimates for LDL-and HDL-cholesterol, triacylglycerols, BMI, WHR, fasting insulin and glucose, systolic and diastolic BP, and CHD from the summary result files of reference GWASs. As some of the reference GWASs were conducted by using cardio-metabochip data, we were able to find 278 SNPs that had effect estimates available for all traits.…”

“…As some of the reference GWASs were conducted by using cardio-metabochip data, we were able to find 278 SNPs that had effect estimates available for all traits. In addition, we removed SNPs based on their high correlation with other SNPs (r 2 > 0.5), as strong linkage disequilibrium (LD) between the SNPs might result in too narrow CIs [13]. This resulted in 257 SNPs.…”

“…The effects of genetic variants on other risk factor traits were taken into account in the modelling so that the SNP effects on other traits (β Trait_2 ,…,β Trait_n ) were included as covariates in the linear regression model. The regression model parameters were estimated using weights defined by inverse standard errors of β CHD to account for uncertainty of the β coefficients [14]. , call rate < 95% (99% for SNPs with minor allele frequency [MAF] < 5%) were excluded.…”